Examining efforts to predict day of surgery cancellation (DOSC): A systematic review protocol

Day-of-surgery cancellation (DOSC) in elective surgery occurs in roughly 18% of elective surgeries worldwide. This impacts patient physical health, psychological wellbeing, and social function. Further impacts include reduced health service efficiency and wider economic productivity. There are a range of contributing variables including patient factors, resource constraints, and health service pressures which could be integrated into predictive models. This article describes the protocol for a systematic review to evaluate peer-reviewed original research articles and implementation studies of models to predict DOSC. Such statistical models could, if properly integrated into clinical practice, yield benefits to patients and healthcare providers. The systematic review will provide a comprehensive synthesis of evidence in this area to inform future efforts at gold-standard statistical modelling. Predictor-finding studies, subsequent publications of the same model and publications in which the predictive variables have not been disclosed will be excluded. Searches will be conducted in Medline, Embase, Scopus and Web of science. Risk of bias will be assessed using the prediction model risk of bias assessment tool (PROBAST). Data will be collected on included variables, method of prediction, whether prediction was made at the level of the patient or the system, and training and assessment processes. This data will be subject to qualitative synthesis and used to generate a narrative summary and figures. This systematic review will abide by the 2020 PRISMA guidelines. This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors. This review is registered on PROSPERO, registration CRD42023478984.


INTRODUCTION
Day-of-surgery cancellation (DOSC) in elective surgery is recognized as a serious problem to both patients and healthcare providers [1].Such cancellations impact on patients through prolonged burden of physical disease, anxiety, and social factors including unnecessary absence from work, childcare and transport.Health services suffer wasted resources, prolonged waiting lists, and poorer clinical outcomes [2,3].A range of variables likely contribute to DOSC, including patient factors (e.g.change in medical condition), logistical challenges (e.g.bed pressures) and health service pressures (e.g.staff shortages), necessitating a systems approach [2,3].A systematic review suggested DOSC occurs in roughly 18% of elective surgeries worldwide, predominantly due to lack of available operating theatres, insufficient staffing, and change in patient condition [4].This represents a significant impact on individual patients, the health system and wider economic productivity.
Predictive models can be embedded into efforts to improve patient care [5].Theoretically DOSC could be predicted by statistical models in which predictive patient and system factors are inputted and either the probability of an individual being cancelled, or the number of cancellations expected at a given point in time calculated.Chosen output would depend on modelling approach, selected variables, data availability, and study aim.Best practice stresses the importance of having a structured approach to identifying relevant explanatory variables before commencing model development [6].Recognising the potential patient and system benefits of accurate prediction of DOSC, this systematic review will evaluate existing examples of such model development [7].The resultant review will permit an assessment of current practice and provide a comprehensive repository of information relating to examined explanatory variables.Predictions may be made at the level of the patient (i.e.what is the chance that this patient experiences DOSC) or the system (i.e.how many patients will be cancelled at this centre on this day), both of which will be evaluated.Arguably, a well performing model that identified either an individual as being at high risk of future cancellation (say on a future day of admission) or would allow a hospital to forecast a day when significant numbers of cancellations were expected, would provide an opportunity for intervention.Patients felt to be at a 'high-risk' of future cancellation could be counselled in advance, or given the opportunity to rearrange their admission, whereas days with high cancellation forecasts could trigger hospitals to open contingency areas or prioritise emergency cases in advance to free-up elective capacity.To realise such ambitions, it is vital that well-performing models are identified (to permit their validation in external settings), and that current knowledge is synthesised to inform future statistical development.
We present our protocol for a systematic review to synthesise current best practice on the prediction of DOSC.The aims of this systematic review are to assess the methods used to predict day of surgery cancellations, the statistical methods employed to achieve this, and the breadth of explored explanatory variables.Our proposed review will be a crucial addition to the literature, summarising efforts to predict this important patient and system outcome across income setting, surgical specialty, and predictive level (patient vs system).By collating all peer-reviewed studies addressing this question into one location, it will also serve as a resource for future prediction model development and validation studies: providing a comprehensive summary of evaluated predictor variables and study settings.Such a resource will be of great value, as accurate prediction of DOSC could yield significant benefit to both patient experience and hospital flow.

METHODS AND ANALYSIS
This systematic review will abide by the PRISMA (Preferred Reporting Items for Systematic Review and Meta-Analysis) guidelines [8].It is registered with PROSPERO, registration CRD42023478984.

Search strategy
The search strategy included papers from electronic databases: Medline (via Ovid), Embase (via Ovid), Scopus and Web of science.These databases were searched in the timeframe from 1 st January 2013 to 13 th October 2023, limited to English language, using the strategy in Table 1.

Inclusion and exclusion criteria
This review will include peer-reviewed journal articles detailing the development, validation, updates and implementation of models to predict DOSC in which the methods, variables and performance are reported.The inclusion and exclusion criteria, detailed in Table 2, were validated in pilot screening of 200 randomly selected studies by TS and confirmed by decision with two other authors (KK, DJS).Crucially, we will only include studies where a formal statistical model is formulated as defined by key components of the PROBAST guideline [7].Broadly we interpret a formal predictive statistical model to comply with the development or validation stages of this document or the derivation and assessment of a model to address a specific research aim [9].We will not include 'predictor-finding' or simple associative studies that do not proceed to a formal statistical synthesis.

Data extraction
Studies will be uploaded to Zotero version 6.0.27 [10] and duplicates removed using inbuilt features.Title and abstract screening will be carried out by 2 independent authors using Rayyan [11].Each abstract will be screened by two reviewers, conflicts will be arbitrated by a third senior author (DJS or KK).Full text screening will be carried out by 2 independent authors using Rayyan, recording reasons for exclusions.Each full text will be screened by two reviewers independently and conflicts resolved by DJS or KK.A PRISMA flowchart will be generated from this process.
Risk of bias will be assessed using the PROBAST tool [7].The results of the risk of bias assessment will be reported for each study and included in supplementary material.Missing results for each parameter of interest will result in the study being excluded from the analysis of that parameter.The studies included and excluded will be reported in the results.Where there is data reported for the parameter but the study states there was missing data, this will also be reported.

Data analysis
Data extraction will be carried out by 2 independent authors.Where the data parameters of interest are stated to have been collected but are not available in the published paper or supplementary material, the study investigators will be sent a request by email.Data will be collected about the method of prediction, level of prediction, cohort characteristics, setting, method of data collection for model training, variables included, model assessment and validation method and performance.Details are outlined in table 3.
Data will be subject to qualitative synthesis.Qualitative synthesis will be structured around the 9 points of the SWiM (synthesis without meta-analysis) guidelines [13].Studies will be sub-grouped for analysis by prediction level (prediction made at the level of the patient vs at the level of the system) and setting (studies in HIC vs LMIC) [12].Analysis will be carried out on all studies in aggregate, in 2 sub-groups by prediction level, in 2 sub-groups by setting and in 4 sub-groups by prediction level and setting.Studies with missing data for the parameter of interest will be excluded from the analysis and those included or excluded stated in the results.Results of synthesis will be reported in a narrative summary and displayed in tables and graphs.This article reports on a protocol for a systematic review to identify studies developing statistical models to predict day of surgery cancellations at both patient and system levels.The resultant review will be a valuable synthesis of current best-practice and could serve as a core reference for emerging studies in this field.Our proposed review involves a comprehensive search of four major literature databases and complies with best practice on dual screening of abstracts and dataextraction.By stratifying identified studies by income setting we hope to be able to provide insights into data variables examined in different income settings that may be driven by potential data availability.As with many systematic reviews we acknowledge the limitations of our proposed search, especially only including studies published in English language [14].The use of multiple reviewers aims to minimise bias.

ETHICS AND DISSEMINATION
No ethical approval is required.
This systematic review will be disseminated in a peer-reviewed journal article.