The study cohorts were derived from a large-scale regional quality initiative in the province of British Columbia, Canada, shortly after the publication of the Canadian Association of Radiologists guidelines for the prevention of CIN which involved the use of a protocol aimed at preventing CIN for clinicians ordering CT scans [9]. Specifically, study cohorts were sampled from four hospitals in the region at two time periods: CT scans performed during December 1–12, 2012 before protocol intervention (Period 1) and CT scans performed during October 1–13, 2013, 10 months after protocol intervention (Period 2). Using the health authority’s radiology software, we were able to identify all patients who received CT scans within the participating hospitals during our study periods, as well as details about their scan types. Patient data from the radiology software was then electronically linked with the health authority’s electronic medical records to collect further information including demographics and laboratory information, matched via each patient’s unique identifier code. Collected data were age, sex, type of CT scan, and sCr measurements taken within a 7-day window prior to and after the CT scan. CT scans were categorized into two types: “Contrast” CT scans included scans with administration of IV or IV and oral contrast medium, whereas “Non-Contrast” CT scans included scans in the absence of any contrast medium or oral contrast medium only. Patients were excluded if their CT scans did not involve the head, spine, chest, abdomen or pelvis (i.e. extremities), or if intra-arterial contrast was used, or if a patient received more than one scan within a 7-day period. This study was approved by the Providence Health Care Research Institute ethics board.
The details of the CIN prevention protocol are available in the Appendix. The evaluation of the protocol was not the purpose of the study, but merely provided an opportunity to collect data on convenient samples to explore the incidence of AKI in hospitalized patients who underwent CT scans.
CIN prevention protocol implementation
Based on current evidence, the CIN prevention protocol was designed to reduce CIN through evidence-based strategies: 1) better surveillance and measurement of serum creatinine pre- and post-contrast scan and 2) mitigate risk by stopping specific medications, and ensuring adequate volume repletion. Radiologists, nephrologists and internists were involved in the development and dissemination of the recommendations. On July 1st, 2013, the protocol was distributed electronically to all staff members within the health authority. At the same time, it became available in paper format throughout every participating hospital within the health authority. There was no pre-printed order, and no electronic decision support tool accompanying the roll out. The use of the protocol was not mandatory in order to order a CT scan, and no mechanism was implemented to track the use of the protocol.
Outcomes of interest
The primary outcome of interest was AKI as defined as per the KDIGO: an acute change in sCr of ≥26 μmol/L within 7 days post-CT scan [5]. The secondary outcome of interest was the availability of sCr assessment before and after CT scan between Period 1 and Period 2, as a measure of ‘awareness of risk’ by clinicians. SCr availability was grouped into three categories: 1) sCr available before and after CT scan, 2) sCr available either before or after CT scan, and 3) no sCr was available before or after CT scan.
Statistical methods
Demographic data were summarized by periods and by CT-scan types. Continuous variables were reported in median with interquartile range, categorical variables in frequency with percentages. Two-way ANOVA and logistic regression model were used to examine differences in baseline variables. Logistic regression model was also used to examine the proportion of patients with AKI for the combinations of periods and CT-scan types, adjusting for age, sex and baseline estimated glomerular filtration rate (eGFR) in the model. Specifically, we were interested in the following four comparisons: (i) Contrast CT-scan vs. Non-contrast CT-scan within Period 1, (ii) Contrast CT-scan vs. Non-contrast CT-scan within Period 2, (iii) Period 2 vs. Period 1 among Contrast CT-scans, and (iv) Period 2 vs. Period 1 among Non-contrast CT-scans. To compare the proportions of sCr availability between periods, we fitted a multinomial logistic regression and examined the difference between periods pertaining to the contrast CT-scans only.
A p-value <0.05 was considered as statistically significant. All statistical analyses were performed in SAS, version 9.3 (SAS institute, Cary, NC). Graphical presentation of the data was created in R, version 3.0.1 (R Foundation for Statistical Computing).