key: cord-0328179-10p02zjl
authors: Rezel-Potts, E.; Gulliford, M. C.
title: Sepsis Recording in Primary Care Electronic Health Records, Linked Hospital Episodes and Mortality Records: Population-based Cohort Study in England
date: 2020-10-14
journal: nan
DOI: 10.1101/2020.10.12.20211326
sha: 140688e8e2446f7dbe369999ebeab7addf414740
doc_id: 328179
cord_uid: 10p02zjl

Objectives: Sepsis is a growing concern for health systems, but the epidemiology of sepsis is poorly characterised. We evaluated sepsis recording across primary care electronic records, hospital episodes and mortality registrations. Methods and Findings: Cohort study including 378 general practices in England from Clinical Practice Research Datalink (CPRD) GOLD database from 2002 to 2017 with 36,209,676 patient-years of follow-up with linked Hospital Episode Statistics (HES) and Office for National Statistics (ONS) mortality registrations. Incident sepsis episodes were identified for each source. Concurrent records from different sources were identified and age-standardised and age-specific incidence rates compared. Logistic regression analysis evaluated associations of gender, age-group, fifth of deprivation and period of diagnosis with concurrent sepsis recording. There were 20,206 first episodes of sepsis from primary care, 20,278 from HES and 13,972 from ONS. There were 4,117 (20%) first HES sepsis events and 2,438 (17%) mortality records concurrent with incident primary care sepsis records within 30 days. Concurrent HES and primary care records of sepsis within 30 days before or after first diagnosis were higher at younger or older ages and for patients with the most recent period of diagnosis with those diagnosed during 2007:2011 less likely to have a concurrent HES record given CPRD compared to those diagnosed during 2012 to 2017 (odd ratio 0.65, 95% confidence interval 0.60 to 0.70). At age 85 and older, primary care incidence was 5.22 per 1,000 patient years (95% CI 1.75 to 11.97) in men and 3.55 (0.87 to 9.58) in women which increased to 10.09 (4.86 to 18.51) for men and 7.22 (2.96 to 14.72) for women after inclusion of all three sources. Conclusion: Explicit recording of sepsis is inconsistent across healthcare sectors with a high proportion of non-concurrent records. Incidence estimates are higher when linked data are analysed.

Sepsis is a growing concern for health systems. In the UK, sepsis is estimated to account for 57 36,900 deaths and 123,000 hospital admissions annually [1] . The Global Burden of Disease 58

Study estimated that there were nearly 50 million incident cases of sepsis worldwide in 2017, 59 with 11 million deaths representing 19·7% of global deaths [2] . The term sepsis was 60 introduced by ancient Greek physicians, but only in recent years has sepsis come to be 61 defined as a syndrome resulting from the interaction between an acute infection and host 62 response leading to new organ dysfunction [3] . Sepsis is an intermediate state that links an 63 infection, or an infection-causing condition, to adverse health outcomes. The term sepsis is 64 now more commonly used than the term 'septicaemia', which refers to blood-stream 65 infection. In the health care systems of high-income countries, records of 'sepsis' have been 66 increasing in both hospital and primary care settings [4] [5] [6] . A study from the U.S. 67

Massachusetts General Hospital [7] found that recording of severe sepsis or septic shock 68 increased by 706% in the decade between 2003 and 2012, while objective markers of 69 severe infection, including positive blood cultures, remained stable or decreased. In a large 70 study from UK primary care the incidence of sepsis diagnoses increased throughout the 71 period from 2002 to 2017, with an especially rapid increase since 2011 [6] . Alongside 72 increasing use of the term sepsis, case definitions have expanded to include patients with 73 evidence of both acute infection and acute organ dysfunction as having 'implicit sepsis' even 74 when sepsis was not explicitly diagnosed [8, 2] . 75 resistance and antimicrobial stewardship, raising safety concerns about the potential for 83 increased rates of serious bacterial infections if antibiotics are not used when needed [11] . 84 85 Electronic health records provide an important data resource for epidemiological research 86 and health surveillance, especially in health care settings. Data linkage provides 87 opportunities to enhance the completeness of ascertainment of health events across health 88 service sectors and population health registries. The advantages of linked records for case 89 ascertainment have been demonstrated for long-term conditions [12] [13] [14] [15] , but research into 90 the use of linked records for the evaluation of infectious diseases has been limited. However, 91 studies investigating the incidence of community-acquired pneumonia indicate that primary 92 care data alone may lead to under-estimation of the burden of infections [16, 17] . 93

This study aimed to exploit data linkage to evaluate the recording of sepsis across primary 95 care electronic health records, hospital episodes and mortality registrations for individual 96 patients registered at general practices in England. We conducted a population-based cohort 97 study to compare simultaneous recording of sepsis in primary care, hospital episodes and 98 mortality data and to estimate the incidence of sepsis from different data sources. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 14, 2020. . https://doi.org/10.1101/2020.10.12.20211326 doi: medRxiv preprint

The study employed the UK Clinical Practice Research Datalink (CPRD) GOLD database. 105

The CPRD GOLD is a primary care database of anonymised electronic health records for 106 general practices in the UK. The high quality of CPRD GOLD data is well-established [18] . 107 CPRD GOLD has a coverage of some 11.3 million patients, including approximately 7% of 108 UK population, of which it is broadly representative in terms of age and sex [19] . Consenting 109 practices in England participate in a data linkage scheme [20] . Approximately 74% of all is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 14, 2020. . https://doi.org/10.1101/2020.10.12.20211326 doi: medRxiv preprint records, or as a primary diagnosis in HES, or sepsis as any mentioned cause of death in 129 mortality records. In UK primary care records, diagnoses recorded at consultations or 130 referrals to or from hospitals were coded, at the time of this study, using Read codes. We 131

identified sepsis records using a list of 77 eligible Read codes. Incident episodes of sepsis in 132 CPRD were recorded using 55 Read codes with four codes accounting for 92% of events, 133

including 'Sepsis' (two codes) (64%), 'Septicaemia' (18%), and 'Urosepsis' (10%). In HES 134 and death registry records, sepsis diagnoses and sepsis deaths were defined using 23 135 ICD10 codes for sepsis. In HES records we evaluated the primary diagnosis, which accounts 136 for the majority of the length of stay of the episode, with other diagnoses being referred to as 137 comorbidities [22] . Incident diagnoses of sepsis in HES were coded with 20 ICD10 codes 138 with three codes accounting for 89% of events, including 'Sepsis, unspecified' (72%), 139

'Sepsis due to other Gram-negative organisms' (13%) and 140

'Sepsis due to Staphylococcus aureus' (5%). In mortality data, we included all mentioned 141 causes of death because sepsis may be part of a sequence of morbid events and not always 142 an underlying cause of death [23] . 'Sepsis, unspecified' accounted for 93% of causes of 143 death among those in the ONS death registry with sepsis as any mentioned cause of death. 144 145

Incident sepsis events were identified for each data source. We calculated person-time at 147 risk from the start to the end of the patient record. Person-time was grouped by gender and 148 age-group from zero to four, five to nine and 10 to 14 and then 10 years age groups up to 85 149 years and over. Incidence and mortality rates were age-standardised using the European 150 standard population for reference. We searched for concurrent events across data sources 151 using a 30-day time-window. We calculated age-specific incidence rates using primary care 152

EHRs and then adding HES records, mortality records or both. Finally, we fitted a logistic 153 regression model to evaluate associations of gender, age-group, fifth of deprivation and 154 . CC-BY 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 14, 2020. . https://doi.org/10.1101/2020.10.12.20211326 doi: medRxiv preprint period of diagnosis with concurrent sepsis recording. All data were analysed in R, version 155 3.6.3. 156 157

In order to consider recurrent sepsis events, we conducted a sensitivity analysis using 159 CPRD and HES where incident events were first sepsis records during each calendar year 160 during the study period. We also evaluated the effect of extending the time-window for 161 concurrent events from 30 to 90 days. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 14, 2020. Table 1 . Each data source showed a slightly higher proportion of females than 172 males (CPRD: 51%; HES: 52%; ONS: 55%). The frequency of sepsis increased with age, 173 with a maximum over 75 years of age. The most deprived IMD quintile had the highest 174

proportion of sepsis cases for each data source for men and women (25% to 26%) and the 175 least deprived quintile consistently had the lowest (15% to 17%). The number of first sepsis is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 14, 2020. . https://doi.org/10.1101/2020.10.12.20211326 doi: medRxiv preprint Annual age-and gender-standardised incidence of sepsis is shown for each data source in 184 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 14, 2020. Table 1 and 216 Supplementary Figures 3 and 4) . 217

Logistic regression analysis of variables associated with concurrent recording in more than 218 one data source for patients with sepsis found that among patients with sepsis events 219 recorded in primary care, concurrent HES records of sepsis within 30 days before or after 220 first diagnosis were higher at younger or older ages and for patients with the most recent Table 3 shows age-specific sepsis incidence rate per 1,000 patient-years from primary care 229

EHRs and for primary care EHRs combined with HES and ONS, stratified. With primary care 230

EHRs alone, sepsis incidence rates increased from 0.19 in males and 0.20 in females per 231 1,000 PY under five years of age, to 5.22 in males and 3.55 in females at 85 years and 232

older. Estimated incidence rates were substantially higher when either HES or mortality 233 records were included. When all three data sources were combined the incidence of sepsis is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint 

This study evaluated sepsis recording in primary care electronic records, hospital episodes 243 and mortality records for a large population registered with general practices in England. We 244 found that, over a 16-year period, a similar number of incident sepsis events were recorded 245 in primary care and hospital records, However, a high proportion of these records were not 246 concurrent across data sources: the majority of sepsis events recorded in primary care were 247 not recorded in hospital episodes, and the majority of hospital episodes were not recorded in 248 primary care. This conclusion held even when events were evaluated over a longer time-249 window or if recurrent as well as incident events were included. There were a smaller 250 number of records of sepsis from mortality registration but a majority of these were not 251 associated with concurrent primary care or hospital records for sepsis. Analyses of 252 associations between concurrent recording between sources and available patient 253 covariates were not consistent across linkages, suggesting that coding variations were 254 largely unexplained. Estimates for age-specific incidence rates may be up to twice as high if 255 linked data sources are employed. 256

Interpretation 258 259

Current estimates suggest that about 50-70% of all sepsis events are community-acquired 260 [24, 25] . Clinical guidelines recommend that patients with suspected sepsis should be 261 referred for management in hospital, which suggests that most patients seen in primary care 262 may be admitted to hospital [1] . However, some patients with sepsis may access hospital 263 services directly without first presenting in primary care. Investigations in hospital may 264 identify underlying causes for sepsis, which might be coded as the reason for admission. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 14, 2020. . https://doi.org/10.1101/2020.10.12.20211326 doi: medRxiv preprint associated with a fatal outcome [26] . It may be expected that a sepsis diagnosis will be 267 communicated to the patient's general practice or, in the event of a fatal outcome, recorded 268 into mortality records. Our results indicate that the process of recording sepsis episodes 269 across different health information systems is highly inconsistent. Health professionals may is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 14, 2020. . https://doi.org/10.1101/2020. 10.12.20211326 doi: medRxiv preprint 

This study is broadly consistent with the growing body of literature that advocates the use of 295 linked data sources [13] [14] [15] [16] . However, it also indicates that stand-alone CPRD data may 296 provide accurate estimates of changes in the burden of sepsis. Millet et al. (2016) found that 297 population-averaged community-acquired pneumonia incidence was 39% higher using the 298 linked data than stand-alone data [16] . It may be that increased awareness and 299 standardisation of detection and recording have prevented such discrepancies being 300 observed for sepsis. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 14, 2020. . https://doi.org/10.1101/2020.10.12.20211326 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 14, 2020. . Table 2 : Age-specific sepsis incidence rate (exact Poisson 95% confidence intervals) per 1,000 patient-years in each CPRD and CPRD combined with HES and ONS, stratified by gender. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 14, 2020. . https://doi.org/10.1101/2020.10.12.20211326 doi: medRxiv preprint

The National Institute of Health and Care Excellence

Global, regional, and national sepsis incidence and mortality, 1990-2017: analysis for the Global Burden of Disease Study

Sepsis and Septic Shock: A History

The Epidemiology of Sepsis in the United States from

Sharp rise in sepsis deaths in the UK

Serious bacterial infections and antibiotic prescribing in primary care: cohort study using electronic health records in the UK

Comparison of trends in sepsis incidence and coding using administrative claims versus objective clinical data

Sepsis and antimicrobial stewardship: two sides of the same coin

Safety of reduced antibiotic prescribing for self-limiting respiratory tract infections in primary care: cohort study using electronic health records

Completeness and diagnostic validity of recording acute myocardial infarction events in primary care, hospital care, disease registry, and national mortality records: cohort study

Defining upper gastrointestinal bleeding from linked primary and secondary care data and the effect on occurrence and 28 day mortality

Searching multiple clinical information systems for longer time periods found more prevalent cases of asthma

Estimating the prevalence of diagnosed diabetes in a health district of Wales: the importance of using primary and secondary care sources of ascertainment with adjustment for death and migration

Improved incidence estimates from linked vs. stand-alone electronic health records

risk factors and burden of community-acquired pneumonia in COPD patients: a population database study using linked health care records

Validation and validity of diagnoses in the General Practice Research Database: a systematic review

Data Resource Profile: Clinical Practice Research Datalink (CPRD)

Medicines & Healthcare Products Regulatory Agency, National Institute for Health Research. CPRD linked data 2020

Department for Communities and Local Government. The English Index of Multiple Deprivation (IMD)

Data Resource Profile: Hospital Episode Statistics Admitted Patient Care (HES APC)

Sepsis-associated mortality in England: an analysis of multiple cause of death data from

Community-, Healthcare-, and Hospital-Acquired Severe Sepsis Hospitalizations in the University Health System Consortium

Epidemiology and economic evaluation of severe sepsis in France: age, severity, infection site, and place of acquisition (community, hospital, or intensive care unit) as determinants of workload and cost

Funding 314The study is funded by the National Institute for Health Research (NIHR) Health Services 315and Delivery Programme (16/116/46 ). ERP and MG were supported by the NIHR Biomedical 316Research Centre at Guy's and St Thomas' Hospitals. The views expressed are those of the 317 authors and not necessarily those of the NHS, the NIHR, or the Department of Health. The 318 funder of the study had no role in study design, data collection, data analysis, data 319 interpretation, or writing of the report. The authors had full access to all the data in the study 320 and both authors shared final responsibility for the decision to submit for publication. 321 322

The authors have no conflicts of interest. 324 325

The study is based in part on data from the Clinical Practice Research Datalink obtained 327 under license from the UK Medicines and Healthcare products Regulatory Agency. However, 328 the interpretation and conclusions contained in this report are those of the authors alone. 329 330

Requests for access to data from the study should be addressed to 332 martin.gulliford@kcl.ac.uk. All proposals requesting data access will need to specify planned 333 uses with approval of the study team and CPRD before data release.