Data Source
This was a registry-based observational study analysing the effect of COVID-19 pandemic using an interrupted time-series framework.
Data were extracted from the national CR clinical registry. As of 2020, the Slovenian cardiac rehabilitation network includes 7 active regional CR centres (affiliated with regional hospitals) providing structured and comprehensive CR programme through 36 reimbursed sessions adhering to a national clinical pathway. Adherence to the clinical pathway is monitored through process and outcomes indicators, which are collected by the National CR Registry; the registry collects data annually, with cardiac rehabilitation reimbursement contingent on regular data reporting.
Data reported include i) patient baseline characteristics (age, sex, presence of dyslipidaemia, arterial hypertension, diabetes mellitus, cerebrovascular disease, peripheral arterial disease, heart failure, atrial fibrillation, chronic obstructive pulmonary disease, depression, and total number of other comorbidities not captured in the aforementioned list), and ii) patient measures at baseline and at completion of the CR programme (smoking status, blood glucose levels, HbA1c in diabetic patients, complete lipid panel, systolic and diastolic blood pressure, body mass index, exercise test parameters, and the health-related quality of life EQ-5D-3L questionnaire).
During the pandemic, all centres closed in March 2020; the majority of centres remained closed throughout 2020 and then inconsistently reopened for short periods of time, mainly due to staff diversion. For the present analysis, therefore, we only report data from centres that continued uninterrupted adapted provision of centre-based CR. The adapted provision was unified across centres and in line with national public health guidance, governmental regulations and professional recommendations (8, 9). Adapted provision included: limiting number of participants per session (max 3, at least 30 m3 per patient), intensified hygienic provision, mandatory ventilation and/or opening of windows, face covering mandates as per government regulations, regular screening and testing when tests became widely available (at no cost for patients), regular informing of patients on COVID-19 related issues (including promotion of public health recommendations).
The study carried out in accordance to relevant guidelines, regulations and legislative frameworks, informed consent was obtained for study participation/at inclusion into the national CR registry. The study was approved by the National Medical Ethics Committee of the Republic of Slovenia (Ministry of Health, Republic of Slovenia), approval no. KME-0120-557/2021/3.
Outcome Variables
We analysed pre- and post-COVID-19 outbreak time trends for relevant CR outcomes: i) exercise training (i.e., percentage of patients improving > 20% on final exercise testing compared to baseline); ii) risk factor control (i.e., percentage of patients with low-density lipoprotein cholesterol [LDL-C] < 1.8 mmol/L, triglyceride levels < 2.3 mmol/L and/or systolic blood pressure < 130 mmHg); iii) secondary prevention (percentage of patients prescribed high-potency statins); iv) lifestyle intervention (percentage of patients with body mass index [BMI] < 30 kg/m2 and percentage of smokers actively smoking at end of CR); and v) health related quality of life (percentage of patients with positive change in health state value for EQ-5D-3L) (11, 12). Relevant CR outcomes were selected from state-of-the-art CR quality indicator frameworks (11, 13–15), in order to i) comprehensively capture all component interventions of CR and ii) to allow for meaningful detection of change/difference. For instance, high-potency statin therapy (defined as either atorvastatin 40–80 mg or rosuvastatin 20–40 mg) was selected because overall statin prescription rates have reached a ceiling point (> 90% of patients prescribed a statin), whereas novel lipid medications have only recently been introduced (< 5% of patients prescribed a PCSK9 inhibitor with expectedly increasing adoption rates). Target LDL-C < 1.8 mmol/L was selected to allow inclusion of patients who completed CR before 2019, when more stringent target levels were introduced (1). Exercise capacity improvement > 20% was selected based on our previous observation that such improvement is meaningful in predicting long-term cardiovascular outcomes (16) and to allow for different individual baseline capacity and different exercise testing modalities across centres. EQ-5D-3L was introduced as a quality measure in 2019, so for the specific health-related quality of life the time series was restricted to 2019–2021 to allow for balanced pre- and post-intervention time points.
Outcomes data were aggregated over monthly time units (i.e., proportion of patients meeting selected outcome standard per month). Data from patients concluding CR between 1 Jan 2018 and 1 Apr 2022 were extracted from the registry, yielding a total of 52 months/52 aggregated data time points (balance 26 pre-/26-postintervention). For health-related quality of life the time series was restricted to 44 months (balance 22 pre-/22-postintervention).
Summary Statistical Analysis
Data were appraised for normality of distribution visually and formally (using the Kolmogorov-Smirnov test for the population sample, n = 1,452 patients, and the Shapiro Wilk test for the aggregated time-series data sample, n = 52 months). Summary descriptive statistics are expressed as means (with standard deviation [SD]) for normally or medians (with interquartile range [IQR]) for non-normally distributed continuous variables, and as numbers (with proportions) for categorical variables. Exploratory pre-post comparisons (before-and-after interruption) were assessed by t-test, Mann–Whitney U test or χ2 tests, as appropriate. Statistical significance was set at two-tailed p < 0.05. Statistical analyses were performed with Stata/IC 14.2 for Mac (StataCorp, College Station, TX, USA).
Interrupted Time Series Analysis
In line with our hypothesis that the pandemic outbreak would cause a shock response (change in level) followed by gradual weaning/recovery (change in trend), data were fitted to segmented regression models for interrupted time-series (10). Model assumptions were assessed with correlograms (autocorrelation and partial autocorrelation function) and residual plots, and refitted/adjusted as appropriate (e.g., differencing for stationarity, dummy variables for seasonality and/or autoregressive orders).
Segmentation was defined prior to analysis; based on epidemiologic data (first COVID-19 case detected in country on 4 March 2020) and government response (epidemic declared and public life restrictions enacted on 12 Mar 2020), 1 March 2020 was selected as the time series breakpoint. Time series interruption yielded balanced 26 time points before and 26 time points after breakpoint (22/22 for health-related quality of life).
Sensitivity analyses were carried out with 1 March 2019 and 1 March 2021 as alternative breaking points (i.e., to validate the causal impact of the COVID-19 pandemic outbreak time on changes in level and trend of CR outcomes). Secondary/exploratory analysis was carried out by adding monthly averages for sex, age, number of comorbidities and number of CR sessions to fitted models.
Estimates are reported as percentage points (rather than decimal absolutes) to provide intuitive appraisal of change — i.e., representing a percentage points difference in CR outcomes after interruption (pandemic outbreak March 2020) compared with expected estimates had the interruption (pandemic) not taken place.