Cardiovascular complications are a leading cause of mortality after non-cardiac surgery.1,2,3,4 Specifically, myocardial injury after non-cardiac surgery (MINS) is the complication with the second highest attributable fraction of death (15.9%) within 30 days of surgery, behind only major bleeding (17.0%).5 Most MINS events are asymptomatic,6 thus routine postoperative monitoring of troponins is required to ensure detection.7 For those patients diagnosed with MINS, early cardiology assessment and subsequent intervention (such as anticoagulants in select patients)8 may reduce the risk of death.9 Nevertheless, an evidence base is still being developed with regards to management of MINS.6

Risk assessment and management of patients with cardiac risk factors undergoing non-cardiac surgery is complex. Therefore, structured guidelines were recommended in 2016 by the Canadian Cardiovascular Society (CCS).6 Unlike previous guidelines,10 which largely focused on preoperative risk stratification using expensive non-invasive cardiac testing, the newest guidelines focus on possibly more cost-effective use of biomarkers for postoperative monitoring of at-risk patients. This permits early identification and management of patients at increased risk of short-term cardiac complications (e.g., routine postoperative troponin I [TnI] testing). Given the burden of perioperative cardiac complications, the implementation of evidence-based guidelines to better identify and treat high-risk patients is a key priority for patients, clinicians, and the healthcare system.

Evidence-based clinical practice guidelines can improve patient care11; however, their implementation in clinical practice is often unpredictable, slow, complex,12,13 and can be associated with significant costs.14,15,16 Moreover, successful implementation of CCS guidelines requires engagement of multiple stakeholders (e.g., cardiology, internal medicine, nursing, surgery, anesthesiology). Given these complexities, real-world evaluation of implementation is necessary to understand rates of adherence, resource utilization, and effects on patient care and outcomes.17,18,19 Without such evidence, physicians may only adopt certain aspects, leading to heterogeneous and uneven levels of care provided to similar patients within an institution.20

To fill this knowledge gap, we performed a quasi-experimental study to evaluate the effect of implementation of the CCS guidelines for perioperative cardiac risk assessment on post-implementation processes and outcomes at a large Canadian academic institution. Adherence to routine troponin monitoring was assessed, as well as healthcare resource utilization, and clinical outcomes.

Methods

Prior to initiation of the study, research ethics approval was obtained from The Ottawa Hospital Research Ethics Board (REB #20180527-01H). This manuscript follows reporting recommendations for interrupted time series.21 The protocol for this study was posted on Open Science Framework.22

Setting and design

We conducted an interrupted time series (ITS) design, evaluating implementation of the CCS guidelines at The Ottawa Hospital (TOH) General Campus (Ottawa, ON, Canada). The Ottawa Hospital is a large, tertiary care, academic institution serving a catchment area of approximately 1.2 million people. Perioperative care is provided for major orthopedic, general surgery, hepatobiliary, oncology, gynecology, urology, and thoracic patients. The study period was seven months (August 2018 to January 2019); the pre-implementation phase (12 weekly timepoints), the implementation phase (four weekly timepoints), and the post-implementation phase (12 weekly timepoints). The implementation period was specified a priori to allow time for the intervention to be fully implemented, and data from those weeks were censored from all analyses.

Participants

The study population consisted of all patients undergoing elective, non-cardiac surgery who met the following criteria: 1) anticipated admission to hospital following surgery, and 2) age ≥ 65 or age 45–64 with a Revised Cardiac Risk Index23 (RCRI) ≥ 1. All non-cardiac surgical patients were screened for eligibility. The standard of care at TOH in the pre-implementation phase included preoperative cardiac risk stratification at the discretion of clinicians, limited TnI and electrocardiogram (ECG) screening (ordered at discretion of clinician), no a priori determined thresholds of TnI elevation to trigger consultations to either internal medicine or cardiology, and no protocolized monitoring, treatment, or follow-up of patients found to have elevated TnI. Patients in the post-implementation period received care according to the CCS guidelines for perioperative cardiac risk assessment as detailed below.

Data sources

Data for this study were obtained in three ways, with consistency across study phases. First, a prospective screening log registry was maintained to identify all participants meeting CCS criteria for screening in the pre- and postoperative phase. Next, labs, interventions, and outcomes were captured from the Ottawa Hospital Data Warehouse (OHDW) by linking our prospective registry to all sources of clinical, laboratory and administrative data at TOH. At TOH, all patients can be linked across OHDW data sources using their hospital unique identifier, allowing perioperative data to be captured from our Surgical Management Information System (Optum, Eden Prairie, MN, USA), labs and tests from the electronic health record (vOacis, Telus, Montreal, QC, Canada) and admission level data from the Canadian Institute for Health Information Discharge Abstract Database (DAD). The OHDW undergoes standardized data normalization, quality controls and audit, and has been used extensively in peer-reviewed research. The OHDW also contains a case costing system, which links financial, clinical, and patient activity information to define costs associated with each patient. Finally, clinical processes (i.e., consultations, cardiac testing) were independently captured using duplicate retrospective chart review by two authors (S.F., M.M.).

Guideline implementation

The pathway of care adopted by our institution is outlined in Fig. 1. Prior to implementation, our multidisciplinary team (anesthesiology, cardiology, general internal medicine, surgery, nursing) identified key stakeholders. Members from these stakeholder groups were interviewed to identify and address anticipated barriers to guideline implementation. A summary of the identified barriers can be found in the Electronic Supplementary Material (ESM; eAppendix Methods).

Fig. 1
figure 1

Pathway of care flowchart

Full size image

Preoperative CCS guideline assessment and testing

Surgical patients at TOH are seen for in-person preoperative assessment (versus a nursing telephone assessment) based on standard screening criteria in a devoted pre-assessment unit. For patients meeting criteria for in-person assessment, our institution did not perform routine natriuretic peptide testing; therefore, risk stratification relied on RCRI criteria (as recommended by the guidelines). New orders sheets were developed and signed for TnI screening (preoperative baseline value, postoperatively within six hours, and then postoperative days [POD] POD 1, 2, and 3) as well as a routine ECG (postoperatively within six hours), where patients met screening criteria based on the CCS guidelines. Care pathways were co-developed with input from physicians (anesthesia, internal medicine, cardiology) and perioperative nurses. During the study period, no other changes were made to perioperative pathways at TOH.

Postanesthesia care unit

Postoperatively, patients meeting screening criteria had a TnI test and an ECG in the postanesthesia care unit (PACU). Patients with a positive TnI or abnormal ECG who were hemodynamically unstable or had other indications for continuous monitoring remained in the PACU, as per usual care. Patients in the PACU with a positive TnI or abnormal ECG who were asymptomatic and stable with no other indication for monitoring were discharged to the ward as per usual standard of care and were followed by internal medicine or cardiology (i.e., TnI and ECG testing in isolation did not impact patient flow or monitoring without clinical indication).

Postoperative ward care

Once on the surgical ward, patients received daily TnI tests until discharge, or until postoperative day (POD) 3, whichever occurred first. If either a PACU or ward TnI test was above normal limits (TnI > 45 ng·mL−1), the patient received either an internal medicine consult (TnI 45–1,000 ng·L−1) or a cardiology consult (TnI > 1,000 ng·L−1); thresholds were established a priori based on multidisciplinary team meetings. Postoperative consults and the postoperative care pathways of the patients were standardized by the cardiology and internal medicine divisions. Consultations included a complete review of systems and investigations to identify underlying causes of TnI elevation.

Formal cardiac testing and follow-up

For patients who had an elevated TnI (TnI > 45 ng·L−1), nuclear stress myocardial perfusion imaging was performed within one week of discharge, and a follow-up appointment with either a cardiologist or internist within two weeks of discharge.

Primary outcome

Our primary outcome was adherence to troponin monitoring in patients meeting screening criteria based on CCS guidelines. Adherence was defined by the proportion of eligible patients that have TnI drawn according to the following criteria: 1) for patients with a length of hospital stay ≥ three days, at least three of their four eligible TnI drawn, 2) for patients with a length of stay of two days, at least two of the three eligible TnI drawn, and 3) for patients with a length of stay of one day, at least one of two eligible TnI drawn. The data for all eligible patients were aggregated on a weekly basis and expressed as a proportion.

Secondary and exploratory outcomes

Our secondary outcome was the adherence to appropriate follow-up care. This was defined by the proportion of patients with positive TnI that had a general internal medicine or cardiology consult in-hospital or as an outpatient within 30 days of discharge. Exploratory outcomes included: 1) utilization of services in TnI positive patients such as echocardiograms, ECGs, nuclear stress myocardial perfusion imaging, angiograms, percutaneous interventions, and cardiac computed tomography, 2) resource-use outcomes (mean weekly hospitalization costs, mean weekly length of stay), and 3) clinical outcomes (weekly proportion with a postoperative complication [using validated patient safety indicators applied to the DAD] and mortality).

Statistical analyses

All analyses were conducted using SAS v9.4 (SAS Institute, Cary, NC, USA). Descriptive statistics were compiled to compare patient characteristics between the pre- and post-implementation phase (t tests for continuous normal variables, Wilcoxon for skewed continuous variables and Chi squared tests for binary and categorical variables).

To test for changes from pre- to post-implementation, we used an ITS design. This allowed us to assess for a change in outcome from pre- to post-implementation while accounting for the underlying temporal trend across the study period. Analysis of all outcomes used linear segmented regression with terms for baseline intercept and baseline slope, as well as terms for change in intercept and change in slope after implementation of the intervention. Our main measure of interest was the total change from pre- to post-implementation, which is the sum of the change in intercept and change in slope. Observations obtained during the four-week implementation period were censored from analysis. Typically, ITS designs require 12 time periods during each phase, with at least 40 observations per period.24 In our study, each period was seven days and we had 12 periods in each phase. The model was estimated using restricted maximum likelihood estimation and accounted for first-order autocorrelation. The effect of the intervention was described as intercepts and slope changes, together with 95% confidence intervals (CI). To test the impact of possible deviations from assumptions of linear regression, we repeated our segmented regression analyses after transformation of weekly values on the natural logarithm scale. As part of a post-hoc analysis, the number needed to evaluate (NNE) was calculated to determine the number of patients needing to have TnI testing to detect one incident of MINS.

Results

Study population

During the 24-week period, 1,421 patients were included in the study (706 during the 12 weeks pre-implementation, and 715 during the 12 weeks post-implementation period). The baseline demographics of the pre- and post-implementation period patients are shown in Table 1. Both groups were similar with respect to age, sex, American Society of Anesthesiologists physical classification, surgical service rendered, duration of surgical procedure, and baseline comorbidities. Orthopedic, urologic, thoracic, and general surgeries accounted for roughly 80% of the procedures performed.

TABLE 1 Baseline demographics of included patients
Full size table

Primary outcome

The results of the ITS analysis are presented in Table 2, while additional TnI testing results are summarized in Table 3. Overall, adherence was 5.8% in the pre-implementation period compared with 69.4% in the post-implementation period, with 95 (13.5%) of pre-implementation patients receiving at least one TnI test, while in the post-implementation phase, 530 (74.1%) patients received at least one TnI test (Fig. 2). Using linear segmented regression, we identified a significant change in total TnI adherence in the post-implementation phase (+67%; 95% CI, 55 to 80; P < 0.0001). Results of the log-transformed analysis are presented in the ESM (eTable 1).

TABLE 2 ITS analysis results
Full size table
TABLE 3 Clinical outcomes
Full size table
Fig. 2
figure 2

Adherence to TnI testing guidelines. The dashed line represents the adjusted regression line. TnI = troponin I

Full size image

Secondary outcome

In the pre-implementation period, 19 patients had elevated TnI levels (2.7%), compared with 64 patients in the post-implementation period (9.0%). In the post-implementation period, 54 patients (84.4%) received follow-up care in accordance with our care pathway, meaning a general internal medicine consult (TnI 45–1,000 ng·L−1) or a cardiology consult (TnI > 1,000 ng·L−1) in-hospital or as an outpatient within 30 days of discharge (Table 4), compared with just eight (42.1%) during the pre-implementation period. Using the TnI-positive data, the NNE was calculated to be 15.9.

TABLE 4 Resource utilization in TnI-positive patients
Full size table

Exploratory outcomes

The number of TnI-positive patients receiving at least one advanced cardiac test (e.g., computed tomography angiogram) also increased significantly in the post-implementation period. In contrast, the number of TnI-positive patients receiving at least one echocardiogram and at least one ECG and the mean number of ECGs per TnI positive patient remained unchanged post-implementation (Table 4). Full results of our exploratory segmented regression analyses are provided in Table 2. We found no evidence of a change in costs, length of stay, or complication rates between phases (ESM eFigs 1–3). Sensitivity analyses using log-transformed weekly rates were consistent with the primary analyses (ESM eTable 1).

Discussion

In a single-centre implementation study, we found increased adherence to the CCS perioperative guidelines recommending postoperative TnI screening using a multidisciplinary engagement strategy. These findings suggest that similar approaches in other centres could increase adherence to best-practice guidelines. Nevertheless, multicentre studies adequately powered for clinical and patient-centred outcomes will be required to evaluate the effectiveness of the CCS guidelines.

Our study is not without limitations. First, our evaluation was limited to a single Canadian centre in the province of Ontario, so the generalizability of our findings to other centres (e.g., non-teaching hospitals and non-Canadian centres) is unknown. Furthermore, while our TnI testing data and length of stay outcome were directly captured from source data, our cost data were limited to in-hospital (excluding physician services) costs only and our complications outcome is not well-suited to identify specific complications types (such as major adverse cardiac events). Furthermore, mortality rates were low, precluding any formal analysis for death as an outcome. The ITS design is considered to have the lowest risk of bias for evaluating health system interventions when randomization isn’t feasible,25 and our analysis was based on data collected in a close temporal period with no measurable differences between pre- and post-implementation populations. Nevertheless, ITS designs can still be subject to biases. Specifically, while we were not able to identify any concurrent changes to care, unidentified temporal changes could influence results and would not be accounted for by our model.

Successful implementation of evidence-based guidelines is a complex and challenging process, which depends on effectively developing strategies to overcome recognized barriers.15,16 At our centre, we involved key stakeholders in a multidisciplinary group (i.e., cardiologists, internists, surgeons, nurses, anesthesiologists, residents, and clinical research staff). We then performed semi-structured interviews with key stakeholder groups to identify barriers, which included organizational constraints, lack of resources, and concerns about increasing workload.26,27,28 This allowed us to design screening forms more efficiently, optimize workflow to minimize burden on clinical staff, and identify areas of particular concern among stakeholder groups (i.e., evening/weekend operations). Perioperative leadership also disseminated information within their departments on multiple occasions to help prepare staff and detail upcoming changes. A similar approach to implementing the CCS guidelines has recently been described.29

Implementation of guidelines often increases inputs of healthcare resources (e.g., testing, evaluation); therefore, guideline implementation can lead to increased resource use if resulting outputs are not improved (e.g., clinical outcomes, patient flow). In our study, we did not find strong evidence that either length of hospitalization or overall costs increased with guideline implementation, despite performing a significantly higher number of postoperative tests, and significantly increasing the number of TnI-positive patients receiving an appropriate consultation. A previous study, based on cost-consequence modelling rather than direct resource use measurement, suggested that the incremental costs associated with a missed MINS event was 1,632 CAD in 2015.30 Although we did not observe cost savings with routine monitoring, our finding that implementation was cost neutral suggests that the perspective of a health economic analysis must be considered. Our data captured hospital costs (therefore, in-hospital tests and diagnostics were included). Nevertheless, any physician services (e.g., consultation with a cardiologist or internist in- or out-of-hospital) and post-discharge interventions were not captured and could have increased costs at the health system level. Future multicentre implementation studies should capture all associated costs. With methods to better identify high-risk patients,31 and emerging new therapies for MINS,8 further benefits to routine testing could be realized. Future multicentre evaluations of CCS guideline implementation adequately powered to capture important differences in clinical and patient-centred outcomes will ultimately be required to determine the effectiveness of CCS guideline-based care. Nevertheless, such studies will be informed by effective implementation strategies developed in studies such as ours.

Conclusion

Our study showed the ability to implement a locally adapted clinical care pathway to increase adherence to the CCS guidelines for perioperative cardiac risk assessment without increasing either length of stay or in-hospital costs. Nevertheless, the clinical impact of the guidelines remains unclear, and large-scale multicentre evaluations of CCS guideline implementation are needed to gain a better understanding of the potential implications on clinically relevant outcomes.