Background:
A steady increase rates of Escherichia coli bacteraemia has been observed across the United Kingdom over the last decade. In England, this resulted in an extension of Public Health England’s mandatory healthcare associated infection surveillance programme to include the collection of all cases of E. coli bacteraemia, from June 2011.
Trends in bacteraemia have been monitored in Wales since 2005, E. coli is the leading causative organism every year. The numbers and rates of E. coli bacteraemia increase year on year. In 2013, the Chief Medical Officer for Wales requested a Root Cause Analysis into the rise of E. coli bacteraemia in Wales.
Aim:
To examine risk factors affecting survival post- E. coli bacteraemia (ECB), a population based record linkage study.
Methods:
1. Study Design: Population-based retrospective cohort using linked health data analysis for public health investigation. Population-based retrospective survival analysis using electronic linkage of anonymised routine microbiological and healthcare usage data.
2. Dataset preparation
• SAIL is an anonymous data linkage system that holds routinely collected data for research
• Anonymised blood and urine data (2005-2011) from Public Health Wales microbiology database included in SAIL
• Encrypted Anonymous Linking Field (ALF_E) allows electronic linkage of microbiology data to other routinely collected administrative datasets held in SAIL.
3. Dataset linkage
• E. coli blood cultures linked at patient-level by ALFe to datasets within the SAIL databank; hospital inpatient, and demographic datasets.
• Demographics analysed include; age, sex, WIMD 2008 fifths and comorbidity score.
• Comorbidity scores generated per patient based on NHS Information Services’ Charlson Comorbidity Index methodology adapted to local coding by Bottle & Aylin (2011) Journal of Clinical Epidemiology 64 (2011) 1426- 1433
4. Data Analysis:
Exposure variable; Exposure was classified by infection source; COCAI (community onset, community acquired), COHCAI (community onset, healthcare associated), HOHCAI (hospital onset, healthcare associated)
Outcome variable; Follow up for survival was initially set to 365 days post blood culture specimen date, which was adjusted if a patient migrated or died during follow up.
Analysis was restricted to first ECB episodes and patients continuously resident in Wales 1 year prior to the blood culture specimen date to assess comorbidities. Follow up was initially set to 365 days post- blood culture, which was adjusted if a patient migrated or died during follow up. Cases with unknown infection source were removed from the analysis (n=9,712).
Kaplan- Meier survival curves, univariate Cox regression and multivariate Cox regression analysis were performed.
Results:
1. Summary statistics- Kaplan Meier survival curves
a) The estimate of the risk of death 30 days, 90 days and 365 days after E. coli bacteraemia is 18.1%, 25.9% and 37.3% respectively with 95% confidence intervals.
b) Survival curves by sex. Males have a significantly greater risk of mortality compared with females. For males, the estimate of the risk of death 30 days, 90 days and 365 days after E. coli bacteraemia is 20.7%, 29.3% and 41.6% respectively. For females, the estimate of the risk of death 30 days, 90 days and 365 days after E. coli bacteraemia is 16%, 23.2% and 33.9% respectively.
c) There is a significant difference in survival time post ECB by age group. As expected, the older the patient, the greater the risk of death. For patients aged less than 50 years, the estimate of the risk of death 30 days, 90 days and 365 days after E. coli bacteraemia is 6.8%, 9.1% and 13.3% respectively. For patients aged between 50 and 64 years, the risk of death 30 days, 90 days and 365 days after E. coli bacteraemia is 15.9%, 21.9% and 31.1% respectively. For patients aged between 65 and 79 years, the risk of death 30 days, 90 days and 365 days after E. coli bacteraemia is 18.5%, 25.3% and 35.9% respectively. For patients aged 80 years or older, the risk of death 30 days, 90 days and 365 days after E. coli bacteraemia is 22.0%, 32.9% and 48.0% respectively.
d) There is a significant difference in survival time post ECB by infection source.
From those with a known source, those with a community onset infection had a much lower risk of death compared to hospital onset infection source. For patients with a hospital onset, healthcare associated infection (HOHCAI), the estimated risk of death at 30 days is 27%, 38.5% at 90 days and 52% at 365 days. Comparatively patients with a community acquired infection (COCAI), the estimate of the risk of death 30 days, 90 days and 365 days after E. coli bacteraemia is 14.4%, 20.8% and 30.8% respectively.
(For patients with a COHCAI, the estimate of the risk of death 30 days, 90 days and 365 days after E. coli bacteraemia is 18%, 25.5% and 40.5% respectively.)
e) There is a significant difference in survival time post ECB by comorbidity index value.
There is a significantly higher risk of death post ECB with a higher comorbidity score. Therefore the sicker the patient, the more likely they are to die post ECB.
For patients with the highest comorbidity score, the estimate of the risk of death is 27.5% at 30 days, 39.5% at 90 days and 57% at 365 days.
Comparatively for those with the lowest comorbidity score, the risk of death is 8% at 30 days, 12% at 90 days and 17% at 365 days after E. coli bacteraemia.
2. Univariate cox regression
Risk factors include; male gender, older age, higher comorbidity, COHCAI and HOHCAI compared to COCAI.
Gender: The hazard of a male dying at any particular time point is 1.29 times the hazard of a female.
Age: the hazard of a patient aged between 50 and 64 years dying at any particular time point is 2.48 times the hazard of a patient aged less than 50 years.
3. Multivariate cox regression
Risk factors include; male gender, older age, higher comorbidity, COHCAI and HOHCAI compared to COCAI. There was significant interaction between age and comorbidity, therefore interaction terms were included in the final model.
While controlling for age and comorbidity; male gender had increased risk (aHR: 1.18; 95%CI:1.11-11.26), as did healthcare associated infection sources, compared to COCAI, (COHCAI aHR: 1.26; 95%CI: 1.14 -1.39; HOHCAI aHR:1.58; 95%CI: 1.47-1.70)
4. Risk profiles
Lowest risk- Female, aged less than 50, no/low comorbidity, community acquired infection.
Highest risk- Male, aged 80 or over, high comorbidity, hospital onset ECB
The hazard of dying at any particular time point for a male , aged 80 or over, with a high comorbidity index value that develops a hospital onset ECB is 36 times the hazard of a female, aged less than 50, with a low comorbidity index value that develops a community onset ECB infection.
Conclusion:
Over a third of E. coli bacteraemia patients died within a year of developing the infection.
The older and sicker the patient is the more likely they will die. Being male and developing hospital-onset E. coli bacteraemia results in a higher likelihood of death.
The increase in E. coli bacteraemia is a growing public health concern; this study identifies risk profiles of patients that are more likely to die within a year.