Bradford is the seventh largest metropolitan district in England and Wales with a population of over half a million. It is ethnically diverse, with 56.7% identifying as of White British origin, and 25.5% identifying as of Pakistani origin. The population consists of 25.8% aged <18 years and 74.2% aged ≥18 years (www.ons.gov.uk).

The study included patients who visited the Bradford Royal Infirmary hospital between 1 January 2018 and 31 August 2021. This was a period in which the command centre was introduced through a number of phases immediately prior to the coronavirus disease (COVID-19) pandemic.

This is a retrospective population-based cohort study undertaken as part of a mixed-methods evaluation project with a formal evaluation protocol published by the authors in January 2022 [13]. A qualitative study of the command centre programme gave rise to two hypothesized intervention timelines, one focusing on the implementation and activation of the technological components of the command centre and the other a ‘complex’ intervention model that sought to account for the broader patient flow and operational redesign programme in which the command centre technology was a part. For the technology model, a three-phase, interrupted time-series model was used to reflect incremental implementation of the visual displays in the command centre, consisting of a pre-intervention (Baseline), first intervention component (‘command centre displays roll-in’), and second intervention component (‘command centre activation’). For the complex intervention model, a five-phase, interrupted time-series model was used that consisted of pre-intervention (Baseline), first intervention component (‘onset of patient flow program’), second intervention component (‘command centre displays roll-in’), third intervention component (‘command centre activation’), and fourth intervention component (‘hospital wide engagement and training’), the latter referring to the roll-out of remote access to command centre data across the hospital. See Fig. 1 for the details of the time interrupts. Note that the COVID-19 pandemic caused serious stress on the hospital, and global health systems, commencing February 2021.

Data from the hospital’s Secondary Use Services data were provided by the Connected Bradford data service [14] and uploaded to a trusted research environment on a Google Cloud Platform. Relevant data were then extracted by one of the authors (T.F.M.) from the Google Cloud Platform.

Public and patient representatives were consulted throughout the study period through workshops at the command centre. Representatives contributed to the development of the research protocol [13] and the selection of indicator outcomes. The lead patient representative (N.S.) has critical input and is a coauthor of this publication.

A range of patient flow and data quality outcome indicators were identified in the study protocol [13] and are listed below.

Dummy variables were created for each of the intervention components—‘onset of patient flow program’, ‘command centre displays roll-in’, ‘command centre activation’, and ‘hospital wide engagement and training’. We also identified the onset of the COVID-19 pandemic and subsequent spikes in its impact [15]. The components of the intervention were given a value of ‘1’ starting from the date of its introduction until the introduction of the next component or phase and then a value of ‘0’ for the rest of the period. The ‘COVID-19 pandemic’ was given a value of ‘0’ through February 2020 and a value of ‘1’ thereafter. A spike dummy variable was also added by setting ‘1’ for the COVID-19 spike periods based on the UK data [15] and ‘0’ throughout.

A continuous incremental time variable was coded from the start of the time series (e.g. 1, 2, 3, 4). The intervention phases were also modelled using five continuous time variables with ‘0’ in the pre-intervention period, ‘1, 2, 3, 4…’ from the onset of the intervention phase until the end of the phase then level-off for the rest of the study. In addition, seasonality was modelled by including dummy variables for the number of weeks in a year.

Interrupted times-series linear regression analysis [16] was used to assess the impact of the command centre on patient flow and data quality measures. First, linear time-series models were fit to the data. Tests for serial autocorrelation test of residuals were conducted and all tests were nonstatistically significant. Hence, regression models with autoregressive integrated moving average errors were not sought. The Akaike information criterion [17] and the Bayesian information criterion [18] were used in selecting the best-fitting models for the data.

To estimate the average transition time between different stages of A&E care and to map the destinations of A&E patients, a process-mining technique was used [19].

A five-phase interrupted time series was used for the main analyses. To explore if the technology alone had an impact on outcomes, a three-phase interrupted time-series model was used as sensitivity analysis. The ‘broad patient flow program’ and ‘hospital wide engagement and training’ were assumed as independent events of the command centre and adjusted as independent dummy variables in sensitivity analysis models. Five percent significance level and 95% confidence intervals (CIs) were adopted throughout. Analyses were implemented in R (Version 4.0.2).

A total of 203 807 inpatients and 197 084 A&E visits were included in the study.

In this pre- and post-intervention comparative study using time-series data from the hospital’s information systems, the findings indicate that the introduction of the command centre, including software technology and process changes, or the software technology alone had no significant and consistent measurable impact on patient flow and data quality. Based on patient flow indicators, the length of stay showed a nonsignificant decrease of 8.8 h [standard error (SE): 4.5], 8.9 h (SE: 4.9), 1.7 h (SE: 4.4), and 0.55 h (SE: 6.8). The waiting time (time taken until patient treatment), clinician seen time (time until patient is seen by a clinician), and A&E transition time (time taken to progress from one stage of A&E care to the next stage of A&E care) also did not significantly improve during the study period.

The data quality also was worse when the command centre was introduced. On average, the proportion of missing treatment dates and clinician seen dates was 42.7% and 23.4%, respectively. In comparison to the pre-intervention period, the proportion of A&E visitor’s missing treatment dates was higher by 3% (SE: 45.8), 10.2% (SE: 6.3), 21.5% (SE: 5.6), and 2.3% (SE: 8.7) when a ‘patient flow program’, ‘command centre display roll-in’, ‘command centre activation’, and ‘hospital wide engagement resumption’ programmes were implemented, respectively.

There is a paucity of studies that investigate the impact of multidepartment hospital–based command centres on patient flow and data quality. A recent study in Saudi Arabia reported that the use of a ‘smart centre’ led to a reduction of intensive care unit lengths of stay by 10% in emergency admissions [7]. The intensive care unit length of stay reported by Alharbi et al. [7] disagrees with the findings of this study. In fact, the average ICU lengths of stay (as calculated separately) for the pre- and post-intervention periods are 91 and 108 h, respectively, which is an increase of 18.7% after the implementation of the command centre. One notable difference between our study and Alharbi et al. [7] is that the Saudi Arabian Command Centre was a national hub and the data used were from the first wave of the COVID-19 pandemic, whereas the Bradford Command Centre was used only in the Bradford Royal Infirmary hospital and we have used >3 years’ worth of data in our analyses.

The Johns Hopkins Hospital in Baltimore, USA also reported reduction in time for ambulance dispatches and bed allocation for emergency department–admitted patients [8, 9]. However, we do not have data in our study to support or refute this claim.

Our study has certain limitations. First, health service delivery was significantly affected by the COVID-19 pandemic resulting in rapid system-wide effects, which may have an impact on the population of patients and capacity management in the hospital. Cancellation and postponement of surgical operations were common due to the reallocation of resources during the peaks of the pandemic. Although we attempted to control for the effects of the pandemic in our time-series models, the proximity of the activation of the command centre with the onset of the pandemic surge makes it difficult to isolate the effect of the intervention.

Another potential limitation of the study concerns the focus of this quantitative evaluation on a small number of outcome indicators for what was a system-wide initiative designed to impact many areas. Although informing our intervention models using qualitative research at the study site was a strength in our design, qualitative investigation additionally revealed the complexity of this type of intervention and the challenges of implementation within a pressured acute care environment. This may have influenced the study outcome in a number of ways. Staff recall of the historical implementation timeline was variable, especially for the piloting and roll-in of intervention components, including organizational in addition to technological elements. There were suggestions that the colocation of staff in the command centre room that preceded the roll-in and activation phase for command centre displays may have already been established and coordinating functions sooner than the intervention timeline suggests, leading to under-specification of our model. When considering the challenges observed in implementing the technological aspects of the intervention, including data quality, there may have been a significant time lag between the activation of components and any impact on patient flow outcomes. Given the complexity of our intervention model, we did not seek to control for lagged effects of intervention implementation (the time it takes for an intervention to start to influence detectable outcomes). Rather, we presumed that the effects of the intervention components were instantaneous. Finally, due to data access limitations, we could not explore all outcomes identified for analysis in our study protocol.

Nonetheless, the strengths of the study are 3-fold. First, we have used a large sample size for the analyses (203 807 in-patient visits and 197 084 A&E visits). Second, the use of electronic health records data minimizes the inherent biases and errors in other types of observational data. Third, we employed a robust quasi-experimental design using repeated time-series measurement.

Currently, data on the impact of command centres on patient flow and data quality are scarce although the use of these technologies has widely been adopted in the developed world. Our study also has not found strong evidence of the positive impact of these novel technologies on patient flow and data quality outcomes. Therefore, further research using data from other hospital organizations that use the technologies is warranted.

The findings of the study suggest that a command centre package that includes the change of process and software technology does not appear to have a consistent positive impact on patient flow and data quality based on the indicators and data we used. Moreover, when only the software technology was assumed as a component of the intervention, no consistent and significant positive impact on patient flow and data quality was observed. Therefore, hospitals considering introducing a command centre should not assume there will automatically be benefits in patient flow and data quality.

Data pertaining to analysis results are fully available in the main text of the manuscript and the supplementary material file.