Communication and Patient Safety in Gynecology and Obstetrics - Study Protocol of an Intervention Study (TeamBaby)

Phase 1: Needs assessment, retrospective cohort study and pre-experimental study

The concept of needs assessment is based on a previous approach, which aims to implement the safety of patient transfers in intensive care units. This should lead to a comprehensive understanding of the beneficial and inhibitory conditions as well as potentials for the implementation of the intervention in the field (e.g., how different communication models competencies can be integrated beyond training into clinical practice; possible options): diaries, regular e-mails with tips on practical application, etc.). Observations for the ethnographic analyses are recorded in a standardized observation protocol, focus groups and interviews are conducted with partially standardized guidelines with at least one person from each occupational group in order to gain as comprehensive an impression as possible of possible resources and barriers for the implementation of the intervention. The data will then be analyzed in terms of content using an inductive approach and used to develop the training. Migration background will be considered exploratively in the data collection and testing as well as participatory in the app development (i.e. it will be ensured that patients and accompanying persons with migration background will also be included in the development). Different communication models competencies and tailoring based on the Health Action Process Approach (HAPA) will also be integrated into the training. A short manual for practical support and written documentation of the training process will be prepared, also to avoid failure of implementation. The online training will be offered in the pre-experimental study to all professionals at both intervention sites. The aim is to investigate the connection between training participation and preventable Adverse Events (pAEs), to observe mechanisms of action and to create a basis for the implementation of phase 2. The study has a one-group, pre-/post-design. The study includes all specialists (e.g. doctors, nurses, midwives, obstetricians) at both locations (full utilization) with a total of N = 140 participants. Specialists are trained in interdisciplinary groups of 10 to 12 participants. All target criteria are shown in Table 4.

Implementation phase: Research questions, methods, study population and target criteria Stage Research question Method Target group/Data source Outcome A What is the prevalence of pAEs? Monitoring protocols and routine data Anonymous routine data of the hospital from 2018, Mixed Methods study Quality indicators: Reporting according to safe communication, pAEs, retrospective cost analyses of pAEs B What are effective interventions to optimize communication in everyday hospital life? Updating the status of applications Pubmed, PsychInfo, Cochrane Database Overview of effective interventions and effect strengths C What is the adherence for equivalent patient safety measures (e.g. hand hygiene)? Questionnaire All employees at both intervention sites Adherence to patient safety measures; relationships between adherence and quality indicators D What are possible resources and barriers for the implementation of an intervention in order to optimize communication in everyday hospital life? Ethnographic observation; guideline-based individual and focus group interviews Physicians, nursing staff, training assistants, psychologists, midwives Classify resources and barriers E What is an appropriate strategy to implement effective interventions to optimize communication? Triangulation of results from previous phases Matching of data from interviews, focus groups and questionnaires with evidence from the literature Implementation strategy, regarding resources and barriers in implementing the intervention to patient safety Note: pAEs = preventable adverse events

Since the aim is to achieve full utilization, no inclusion and exclusion criteria are defined. Differences between the measurement times are checked using linear and general mixed models, initial values of the first measurement time are taken into account as covariates. Clinics are modeled as a fixed effect. For the entire analysis, corresponding confidence intervals of 95 % are determined. To calculate the standardized mean difference (SMD) between the two measurement points, the non-standardized regression coefficients are standardized to the pooled standard deviation of the observed measured values. According to Cohen's recommendations, SMDs are interpreted as follows: small effect by SMD = 0.2; medium effect by SMD = 0.5; large effect by SMD = 0.8. The odds ratio of binary outcomes is subjected to a logit transformation to calculate effect strengths. To check whether the intervention effect on the primary target criterion is moderated by external variables, reciprocal interaction terms are considered in the model estimation. Continuous moderators are standardized by z-transformation. The estimated parameter of the interaction between the z-transformed continuous variable and the effect estimator of the group difference thus represents the additional intervention effect if the potential moderator increases by a standard deviation. Participants are only included in the analysis if they participate in the corresponding follow-up online survey and are analyzed according to intention to treat. Differences are interpreted as significant when the p-value of a two-sided test is <0.05.

Phase 2: Quasi-experimental efficacy study

In Phase 2, the effectiveness of the communication intervention for the target group of women giving birth/family members is assessed using the gold standard, i.e. randomized control trial (RCT). In each institution, four employees are enabled to train patients and their relatives on the short manual developed to teach different communication models competencies. When a woman gives birth and her relatives are admitted, they are randomly assigned to a training group (IG) or a control group (CG). The pAEs are collected on a person-specific basis (anonymized) and coded in parallel. With an alpha error of 0.05 and a power of 80%, a sample of N = 352 is necessary (to be analyzed IG: N = 176; CG: N = 176) in order to be able to prove an effect of d = 0.3. With an assumed drop-out of 20%, N = 424 participants (to be recruited IG: N = 212; CG: N = 212) would have to be recruited (recruitment sample). The target criteria are evaluated statistically and clinically as in phase 1. Group differences are checked by means of linear and general mixed models, initial values of the first measurement point are considered as covariates. Clinics are modelled as fixed effects. For the entire analysis, corresponding confidence intervals of 95 % are determined. To calculate the standardized mean difference (SMD) between the two measurement points, the non-standardized regression coefficients are standardized to the pooled standard deviation of the observed measured values. According to Cohen's recommendations, SMDs are interpreted as follows: small effect by SMD = 0.2; medium effect by SMD = 0.5; large effect by SMD = 0.8. The odds ratio of binary outcomes is subjected to a logit transformation to calculate effect strengths. To check whether the primary target criterion is moderated by external variables, reciprocal interaction terms are considered in the model estimation. Continuous moderators are standardized by z-transformation. The estimated parameter of the interaction between the z-transformed continuous variable and the effect estimator of the group difference thus represents the additional intervention effect if the potential moderator increases by one standard deviation. Participants are only included in the analysis if they participate in the corresponding follow-up online survey and are analyzed according to intention to treat. Differences are interpreted as significant when the p-value of a two-sided test is <0.05.

Phase 3: Implementation of the app and case-control study

From all findings in phases 1 and 2, the digital app will be developed in a participatory and theory-driven way. The aim is to determine the extent to which the app supports secure communication processes among specialists and patients/relatives in their daily work and thus leads to a reduction in pAEs. The app helps users (skilled workers, mothers and relatives) to prepare for the interview and to cope with possible problems during the interview. Subsequently, a monitoring takes place in which communication with focus on (1) the own role, (2) the role of the communication partner as well as (3) the resonance (feeling of the uniform understanding) is evaluated. In this way, dyadic self-reported and externally reported data are collected and learning from the communication of all participants is ensured. Controls form the data of the participants of both intervention arms of the preceding RCT study from phase 2 (historical control groups; No-Treatment CG analysis: N = 176; No-Treatment CG recruited: N = 212; Treatment CG analysis: N = 176; Treatment CG recruited: N = 212). This allows a first limited statement about the possible effect of a process supporting digital app. With an alpha error of 0.05 and a power of 80%, a sample of N = 352 (analysis sample) is required to be able to prove an effect of d = 0.3. Due to the historical control group, only N = 176 participants need to be analyzed for the intervention group (IG analysis). Assuming a drop-out of 20%, N = 424 participants would have to be recruited (recruitment sample). Due to the historical control group only N = 212 participants have to be recruited for the intervention group (IGrecruited). The target criteria (primary: pAEs; Secondary: pAE costs, different communication models competencies, training-proximal outcomes such as outcome expectations, behavioral intention, action and coping planning, behavior, self-efficacy expectation, adherence safety culture, perceived safety culture, treatment satisfaction; covariates: sociodemographic characteristics incl. migration status) are evaluated statistically and clinically as in Phase 1 and 2. Group differences are checked by means of linear and general mixed models, initial values of the first measurement point are considered as covariates. Clinics are modeled as fixed effects. For the entire analysis, corresponding confidence intervals of 95 % are determined. To calculate the standardized mean difference (SMD) between the two measurement points, the non-standardized regression coefficients are standardized to the pooled standard deviation of the observed measured values. According to Cohen's recommendations, SMDs are interpreted as follows: small effect by SMD = 0.2; medium effect by SMD = 0.5; large effect by SMD = 0.8. The odds ratio of binary outcomes is subjected to a logit transformation to calculate effect strengths. To check whether the primary target criterion is moderated by external variables, reciprocal interaction terms are considered in the model estimation. Continuous moderators are standardized by z-transformation. The estimated parameter of the interaction between the z-transformed continuous variable and the effect estimator of the group difference thus represents the additional intervention effect if the potential moderator increases by one standard deviation. Participants are only included in the analysis if they participate in the corresponding follow-up survey and are analyzed according to intention to treat. Differences are interpreted as significant when the p-value of a two-sided test is <0.05.