Viewpoint: The clinical and health inequality risk factors for sepsis

Sepsis occurs when the body goes into overdrive to try and combat an infection. Over 11m people globally die of it each year. There is evidence from scientific studies about the risk factors for the development of sepsis, but this evidence is mostly based on hospital-based US studies.

Our project, conducted by the University of Manchester in collaboration with the UK Health Security Agency, aimed to better understand the interplay between clinical risk factors and health inequalities and sepsis morbidity, mortality, and management among diverse patient groups.

What was covered by our project?

The project included a rapid review of literature, data analyses of two large research databases (OpenSAFELY and CPRD) and a discussion of the possible policy implication of the research findings. The databases included anonymised patient data from primary care linked to hospital records (with details on diagnoses in the hospital).

The studies included two large case-control studies in England; a case-control study compares cases (with sepsis) to matched controls (without sepsis) and analyses the differences in the frequencies of risk factors such as deprivation or ethnicity.

Over 330,000 sepsis cases (matched to controls) were analysed; most of these were community-acquired (the sepsis developed following an infection outside the hospital). Patients with sepsis related to a COVID-19 infection were excluded.

The rapid literature review identified about 50 studies. We found that there was a limited number of studies directly investigating the correlation between health inequalities, including factors such as deprivation and ethnicity. Moreover, many of the studies were small-scale and lacked uniformity in defining health inequalities. The review identified a need for studies conducted across more diverse setting and countries, including England.

What did we find?

Increased frequencies of sepsis were found in patients with largest socioeconomic deprivation, frailty and chronic conditions, while patients with ethnicities other than White (including Black, Indian, Pakistani, Chinese, other Asian and mixed) did not show any increases in sepsis risk as a group overall.

We also looked at how sepsis affected patient groups with Core20PLUS5 characteristics – which is NHS England’s approach for reducing health inequalities.

In an analysis of 60 different clinical characteristics, it was found that patients with largest socioeconomic deprivation, when compared with non-deprived patients, suffered more frequently from alcohol problems, COPD, learning disabilities of severe mental illness. These were observed to be all strong risk factors for the development of community-acquired sepsis.

People with ethnicities other than White had higher prevalence of anaemia, diabetes mellitus, chronic liver disease, severe mental health illness and visual impairment, which were all found to be strong risk factors for sepsis.

In severely frail patients, the prevalence of anaemia, chronic heart diseases (such as atrial fibrillation, heart failure, ischaemic heart diseases) and Parkinson disease was substantially higher compared to non-frail patients. Again these are all strong risk factors for sepsis.

Risk prediction models differentiating community-acquired sepsis cases from controls achieved good discrimination (c-statistics between 0.75-0.80) using information on age, sex, chronic diseases, deprivation status, distant history of antibiotic prescribing and learning disabilities, despite not including indicators of the infection severity.

Distant history of more frequent antibiotic prescribing was associated with increased risks of developing sepsis (with higher risks in patients with more use). The fatality rates of community-acquired sepsis cases were found to be high and increased in patients with highest deprivation and severe frailty. However, case fatality rates almost halved over the last 20 years.

What does this mean for primary care?

The authors of this work – with expertise across data science, primary care, sepsis and infection management – have reflected on the findings in this project and literature. Recently, the UK Academy of Medical Royal Colleges has issued recommendations on early diagnosis in healthcare settings and initial treatment of sepsis but without recommendations for the prevention of sepsis in primary care.

Antibiotics are often effective in preventing infection-related complications such as sepsis but are currently prescribed in primary care without considering the patient’s risk of severe complications to the infection. However, antibiotic prescribing in primary care is currently not ‘risk-based’ (which means that patients with very low risks of developing these complications are as likely to be offered an antibiotic as those with higher risks).

Furthermore, better targeting and risk-based prescribing of antibiotics for common infections may be needed especially because antibiotics may increase a patient’s risk of harbouring bacteria that develop antibiotic resistance.

There is increasing evidence that repeated antibiotic use may possibly reduce the effectiveness of antibiotics and increase susceptibility to infections. In other words, antibiotics may have an immediate beneficial effect in treating the bacterial infection but at the risk of no longer being effective in the future; effect now and risk later.

This evidence has come from observational studies and studies of the microbiota (bacteria and other microorganisms normally present in e.g. the gut and throat) and, thus, the causal role of antibiotics cannot be established with certainty. On the other hand, there is no evidence to support the effectiveness of the wide-spread practice of prescribing antibiotic use for common infection repeatedly.

The authors propose the following policy priorities in the prevention of sepsis in primary care, taking into account health inequalities:

1. Focus on risk-based prescribing of antibiotics for self-limiting bacterial infections with implementation of risk prediction models combining infection severity with age, sex, chronic diseases and indicators of poorer healthcare access and worse health.
2. Inclusion of common infection management into the QOF or an incentive scheme of some kind for practices with practice populations with higher prevalence of indicators of poorer healthcare access and worse health.
3. Remote monitoring of worsening of common infections in patients at higher risk of infection-related complications.
4. Strengthening of NICE treatment guidelines for common infections by providing advice on frequently encountered clinical challenges (such as many patients receiving multiple courses of antibiotics over time).
5. Implementation of the Learning Healthcare System for common infections, with clinicians provided with tailored feedback information about the challenges in their practices as well as personalised information for patient and clinician during consultation to support shared decision-making. There is an example of a project that could help with this here – a study we worked on looking at how this kind of system can help in practice is published in the BMJ Open.

Professor Tjeerd van Staa is professor of health eResearch at Manchester University