Patients known to be colonised with CPE are frequently re-admitted to different health services to the one they are diagnosed in. However, using survival analysis, we have shown that this frequency of readmission is not significantly higher than other healthcare system patients, after controlling for comorbidities.
We have also shown that majority of CPE colonisation identification and re-admission occurs in public, metropolitan hospitals in Victoria. This is perhaps correlated with hospital capability and centrality.
One unexpected trend observed in this data is the bias toward male patients. This trend hasn’t previously been reported, and further investigation as to whether this is a health service coding issue or an actual trend is warranted to further the epidemiologic understanding.
This work uses administrative data on patient admissions and separations across Victoria, which includes a large amount of quality assurance testing and input restrictions. By combining with notifiable disease information, which is legally required to be reported upon detection, this analysis uses very high quality input data. By using statistical linkage, we were able to follow patient journeys through the healthcare system, which is highly valuable when considering patient hospital admission risk.
Testing for CPE infection is based on highly varied screening procedures, with increased screening being performed generally in intensive care units, and after an outbreak is detected [20]. As a result, it is unlikely that detection of CPE infection is perfect. This means that potentially individuals who are CPE colonised (and therefore a potential source of transmission) may not be included in this analysis, and these patients may have different hospital admission risk.
Most studies of CPE colonisation are prospective studies [21, 22] that are focussed in one facility or ward. By using routinely reported data, we are able to get a greater understanding of the breadth of patients who may be admitted while colonised with CPE, and also understand their future health seeking behaviour. However, a key limitation of retrospective studies is the inability to change sampling strategies, or investigate trends further than what is present in the data.
The linked nature of the data used in this study is another key strength. The impact of inter-facility patient transfer has been studied previously for CPE [13] and other hospital associated infections [21]. These studies have only been able to draw associations between hospital incidence and patient admissions, while this data allowed us to look explicitly at the patterns of those colonised with CPE.
We focussed on CPE cases (infection or colonisation detection) that had a date of onset while hospitalised. While nosocomial transmission is likely to be the primary source of infection, community and household based transmission has been quantified previously [24]. We note that almost 16% of the patients in the notifiable disease database do not have an associated hospital admission. Many of these are likely returning travellers, but it is possible that community transmission could be a contributor to overall disease burden.
There is a potential impact of censoring of admissions of patients in the CPE positive cohort. As the data covers a fixed period (independent of date of CPE notification), those diagnosed late in the data period may not have had a chance to re-present to hospital. While there is a delay between notification of a new CPE case and their calculated onset date, this delay is small (median 5 days, IQR 1–5 days), and so the downward trend in CPE notifications at the end of the data period is likely to reflect the local epidemiology at the time.
Among the patients who are not readmitted within the study period, we assumed that the reasons are similar between the CPE-positive and CPE-negative cohorts, i.e. the combination of hospitalisation not required, moving out of Victoria, or death. It is possible that the reasons for this lack of presentation differ between population groups. Investigating these potential differences in cohorts requires linkage with an additional dataset (the Victorian Death Index), and is a direction of future work.
During this analysis, we excluded subsequent admissions to the health service where the index admission occurred. This is because we assumed that the health service where CPE was detected would have a process for recording the CPE-positive status of patients and, as a result, would take proactive measures to prevent onward transmission. It is not possible to verify this assumption with this dataset, as information about what type of room and use of transmission-based precautions was not available. Given the focus on detection and management of CPE cases in this region, the assumption is likely to be reasonable in this cohort.
Our description of where CPE cases are detected (i.e. predominantly metropolitan, public health services) may be subject to detection bias, as we can’t exclude the possibility that these health services are undertaking more active surveillance than other services. Nevertheless, we suggest that there is benefit to describing the current pattern of patient flow among known CPE cases.
While we have shown that patients colonised with CPE are frequently transferred around the hospital system, we have not evaluated the impact this may have on transmission. Dynamic models of disease and its control are required for this purpose, and are a direction of future work. We have also not been able to identify the frequency with which patients are appropriately placed in contact precautions after CPE diagnosis when admitted to a different health service.
Patients colonised with CPE are not fundamentally different in their hospital admission risk than those who are similarly ill. From an infection prevention perspective, this means that there is still significant infection risk from the transfer of patients who are known to be positive. This risk results in ample opportunity for between-facility transmission shortly after detection. This supports the potential utility of a centralised alert system to identify CPE positive patients on hospital presentation in Australia.
Map of hospitals in Victoria, Australia, coloured by whether the hospital is a public or private hospital. (a) All of Victoria (b) Greater Melbourne metropolitan area
Add Comment