WHAT IS ALREADY KNOWN ON THIS TOPIC

There are some previous studies on the risks for hospital admission in mpox cases pre-2022 and some early in the global outbreak.

Published descriptions of the impact of mpox cases on healthcare services are limited both pre-2022 and during the global outbreak.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

The findings of this study can be used to inform the public health response and guidance for potential future similar sexually transmitted infections outbreaks among the gay, bisexual or other men who have sex with men population. The findings on the impact on healthcare services can also be used by commissioners of sexual and healthcare services to review their plans for healthcare provision.

Background

Prior to 2022, cases of mpox had been largely limited to previously endemic countries in Central and West Africa.1–3 In May 2022 a global outbreak of mpox (predominately clade IIb B.1 lineage) was identified predominately in gay, bisexual or other men who have sex with men (GBMSM).4 5 Transmission appeared largely to be through sexual networks. By the end of 2022, over 80 000 mpox cases had been diagnosed globally.6

Prior knowledge based on mpox cases prior to 2022 could not be directly applied to the 2022 mpox outbreak. The 2022 outbreak was clade IIb which causes less severe disease than mpox endemically circulating in West and Central Africa pre-2022.7 8 Clinical symptoms in the 2022 outbreak differed with limited spread of the rash, and either no or reduced systemic prodromal symptoms.9–12 Many pre-2022 cases in endemic nations were paediatric or immunocompromised, including undiagnosed or undertreated people living with HIV. Comparatively, the 2022 outbreak affected mostly healthier GBMSM adult populations in high-income nations where treatment for people with HIV is good. The differential impact is illustrated by case fatality rates which were up to 11% prior to 2022 and around 0.1% during the 2022 outbreak.13

In the 2022 outbreak, 1–13% of patients required hospital admission in Europe.13 Reasons for admission included pain management, abscesses, secondary bacterial infection and pain on swallowing.10 12–15 A study across 528 cases in 16 countries found anorectal pain to be the main risk factor for admission.9 Another study of 156 mpox cases in the UK to August 2022 reported severe rectal pain as the most common reason (26%) for hospital admission with 90% of cases identified as GBMSM, and 30% as people with HIV, with well-controlled CD4 counts.16 HIV status had no impact on clinical presentation in other studies from Spain and Germany.13 17 18 While one study found that patients with high CD4 counts had similar outcomes in the 2022 outbreak to those without HIV, another study found patients with undiagnosed, untreated HIV and low CD4 counts were at risk of more severe disease and outcomes.15 18

England experienced a large mpox outbreak with 3553 cases reported between 6 May and 31 December 2022.4 Given the considerable impact on sexual health and healthcare services, we undertook a study to assess the impact on healthcare utilisation and to explore demographic and self-reported behavioural factors associated with emergency department (ED) attendance or inpatient admission. We hypothesised that mpox cases taking pre-exposure prophylaxis (PrEP) were likely to have a higher risk of admission compared with those who did not take PrEP. We used PrEP, taken to prevent HIV infection, as a proxy marker of higher risk behaviour, for example, condomless anal intercourse (CAI).19 20 CAI is a risk factor for sexually transmitted infections (STIs), including herpes simplex virus infection which can cause reduced skin integrity.21 22 Reduced skin integrity has been recognised as a risk factor for clinically severe mpox.23

Methods

In England, clinical or laboratory-confirmed mpox infection must be notified to the UK Health Security Agency (UKHSA).24 After deduplication, demographic and test result details of all laboratory-confirmed cases in England were held centrally in a secure database. To inform decision-making during the outbreak case data was linked to Emergency Care Data Set (ECDS) and Secondary Uses Services data sets (SUS) which include information on A&E attendances and inpatient admissions in NHS hospitals in England, respectively. These data sets are made available to UKHSA under a data sharing agreement.25 26 Case data were cleaned, processed and linked to ECDS and SUS records using a unique identifier (National Health Service or hospital number) and existing standardised methods.27 28 After obtaining ethical approval, the study team was given access to an anonymised version of the data set for analysis.

A hospital episode was defined as a single continuous instance of hospital ED attendance or inpatient admission until a discharge event; patients could have multiple episodes. Patients without a unique identifier required for linkage were categorised as not linked. Those with a unique identifier but no corresponding ECDS or SUS record were categorised as having no hospital usage. Patients with an ECDS but no SUS record were counted as ED attendance. ECDS records were joined to SUS inpatient records as a single record when ED attendances ended with a discharge date matched to inpatient admission at the same hospital provider. Patients with an SUS record were counted as an inpatient admission regardless of whether they were admitted directly or via ED.

Patients may attend hospital for various reasons therefore, analysis was restricted to episodes starting up to 7 days before and within 21 days of a positive laboratory test for mpox. For inpatient admissions, we also included episodes that specifically mentioned mpox in International Classification of Diseases 10th Revision (ICD-10) codes at the time of discharge. We described the number of ED attendances and inpatient admissions as episode counts for all patients. We categorised age in four groups (18–29, 30–39, 40–49 and 50+ years) and region to four categories (London, Midlands and East of England, South of England and North of England) to ensure sufficient numbers in all categories. We created a variable for the month of admission from May to August individually but grouped September to December as a single category. Length of stay was calculated for all admitted patients in calendar days.

For the case–control study, all patients were categorised into one of three categories—no hospital usage, attended ED alone or admitted for inpatient care (with or without ED attendance). For patients with multiple ED attendances or inpatient admissions, the chosen hospital outcome was inpatient admission (if admitted), ED attendance (if attended ED) and no hospital usage in order of priority. The hospital record closest to the date of a positive test was chosen. This data set included information obtained from enhanced surveillance questionnaires on symptoms, sexual behaviours and medical history. Questionnaires were completed voluntarily by cases directly or by phone interviews conducted by trained UKHSA staff.

After checks on data quality and missing data, we conducted univariable multinomial logistic regression comparing the risk of inpatient admission or ED attendance to no hospital attendance as the control. Explanatory variables were age group, region, month of admission, presence of individual symptoms, behavioural factors such as number of sexual partners in previous 12 months, medical history such as number of STI in previous 3 months and current use of PrEP. P values were calculated using likelihood ratio test (LRT). Plausible variables with a p value of 0.2 or less were included in the multivariable model to estimate adjusted relative risk ratios (aRRRs) for multinomial and adjusted ORs (aORs) for binomial logistic regression with 95% CIs.

As a sensitivity analysis, we conducted binary logistic regression comparing the risk of inpatient admission with no hospital attendance. For this analysis, we removed patients who attended ED to ensure that the estimates for hospital admission can be interpreted in relation to patients managed in the community. As post hoc analysis, we hypothesised that patients taking antiretroviral therapy (ART), either for HIV infection or as PrEP, may have similar risk profile and behaviours and hence compared their risk to those not taking ART. Hence, we repeated both multinomial and binary logistic regression by combining individuals taking any ART, either for HIV infection or for PrEP, as a single group and compared their risk to patients not taking any ART. All analysis was conducted in R V.4.2.2. Multinomial regression was conducted using multinom function from the nnet package.

Results

In total, 3542 adult males aged over 18 had laboratory-confirmed mpox infection between 18 May and 24 December 2022 in England. A total of 394 (11.1%) patients could not be linked to their hospital records, of which London accounted for the majority (88.6%). London accounted for 65.7% of linked cases, 71.9% of ED attendances and 59.9% of patients requiring inpatient admissions in England. Key characteristics of the cases, including numbers who could not be linked, are shown in table 1.

Table 1

Key characteristics of adult male patients with mpox infection, England, 2022 (n=3542)

Among 3148 cases with linked healthcare episode usage records, 544 (17.3%) accounted for 688 ED attendance episodes; 430 (79%) had a single ED attendance episode, 89 (16.4%) attended ED twice and the remaining 25 (4.6%) three to five times. Similarly, 202 (6.4%) patients accounted for 328 inpatient admission episodes; 116 (57.4%) had a single admission episode, 61 (30.2%) had two admissions and the remaining 25 (12.3%) accounted for three to eight admissions each. All ED attendances and inpatient admissions occurred between 6 May and 23 November 2022. Distribution of cases by laboratory report date closely mirrored inpatient admission and ED attendances (online supplemental file 1).

Counts and proportions of ED attendances and inpatient admissions by age group, period and region is shown in table 2. Among those that could be linked to hospital records, total bed days for admitted patients was 1285 days, of which London accounted for 796 days (62%). The median length of a hospital admission episode was 3 days (IQR 0–8 days) and 162 (49.4%) patients stayed less than 1 day. Patients younger than 40 years of age accounted for the majority of mpox cases, ED attendances, inpatient admissions and hospital bed days. The peak of ED attendances and inpatient admissions occurred in June to July 2022.

Table 2

Counts and proportion of ED attendances and inpatient admissions among adult male patients with mpox, England, 2022

For the case–control study, 689 adult men were included. Of these, 73 had an inpatient admission, 132 attended ED only and 484 had no hospital usage recorded. Characteristics, some symptoms and the risk factors collected on the questionnaire are shown in table 3. This table also includes LRT p values derived from univariable multinomial and logistic regression.

Table 3

Characteristics of patients included in the multinomial and case–control study (n=689) (19.4% of all adult mpox cases)

For variables in the case–control study, vaccination status was missing for 136 (18.7%) patients, PrEP status for 44 (6%), number of partners for 25 (3.4%), number of STIs for 18 (2.5%) and ethnicity for 15 (2.1%). The presence of neurological symptoms and abscess/swelling were excluded from explanatory variables as less than seven patients (1%) reported such symptoms.

In the multinomial regression, seven variables from univariable analysis with an LRT p value of <0.2 were included in the multivariable model. Results of the model as aRRRs and 95% CIs for attending ED and admission to hospital as compared with not using healthcare services are shown in table 4. Similarly, five variables (the same included in the multinomial regression with the exception of rash and fatigue) with LRT p value of <0.2 in the univariable logistic regression were included in multivariable model. The aORs for this model are found in online supplemental file 2.

Table 4

Multivariable multinomial model for attending ED and inpatient admission, adult male patients with mpox, England, 2022 (n=652)

In the post hoc analyses (online supplemental files 3 and 4) using multinomial regression, individuals not taking ARTs had aRRR of 2.15 (1.22–3.78) for hospital admission and aRRR of 1.13 (0.69–1.85) for ED attendance compared with those taking ARTs. In the binary logistic regression, the aOR of hospital admission was 2.19 (1.24–3.81) for individuals not taking ART compared with those taking ART.

Discussion

This study linking data on mpox cases with hospitalisation records found 3542 adult males diagnosed with mpox in England between May and December 2022, 544 (15.4%) attended ED and 202 (5.7%) were admitted to hospital. This created a considerable impact on healthcare services, particularly in London which saw the most cases. No one specific symptom was seen to impact on ED attendance or inpatient admission. In multinomial regression, we also found strong evidence that compared with people living with HIV, the aRRR for hospital admissions for those not using PrEP was 6.9 (95% CI 2.3 to 20.6) versus 4.9 (95% CI 1.7 to 14.1) for PrEP users.

Number of cases and hospital usage peaked in June to July 2022, however, proportions going to hospital and length of stay decreased after the 23 May 2022 when national guidance was issued advising home isolation of patients where clinically appropriate. Given initial uncertainty around the prognosis and clinical management for individual patients as well as the perceived risk of transmission to the wider community, cases were more likely to have attended and been admitted to the hospital in May 2022. Further guidance was then issued in July 2022 derogating clade IIb mpox as a high consequence infectious disease, ending the need for cases to be admitted for infection prevention control reasons. This may have further added to the fall in the proportion of cases admitted after July.

London experienced the highest burden of both cases and hospital usage as measured by a number of ED attendances, inpatient admissions and bed days. This is unsurprising given that in England, London residents are most likely to report being GBMSM.29 While the numbers and proportions attending ED and admitted to hospital varied somewhat between regions, we found little evidence for an association between region and healthcare usage in the case–control study.

Adult males under 40 years accounted for the majority of cases, ED attendance and inpatient admission. However, the effect of age on ED attendance and inpatient admission was not consistent. In multinomial regression, adults in the 40–49 and 50+ age group had a lower risk of ED attendance compared with those 18–29 years. For inpatient admission, adults in the 40–49 age group had a lower risk compared with the 18–29 age group in both logistic and multinomial regression. Given the moderate numbers of ED attendances and inpatient admissions, it is difficult to be confident of the effect of age on the risk of ED attendance and inpatient admission.

In both multinomial and binomial logistic regression, we found strong evidence that compared with people living with HIV (who are one group ineligible for PrEP), those who were not taking PrEP had seven times higher risk of inpatient admission whereas those taking PrEP had five times higher risk. This effect was not seen for ED attendance. Similarly, our post hoc analysis showed that individuals not taking ART had a higher risk of hospital admission compared with those taking ART, either for HIV treatment or as PrEP. However, our study could not account for potential differences in healthcare-seeking behaviours of people with HIV or taking PrEP. People living with HIV and using PrEP may seek care through HIV and sexual health services already used. These services, compared with ED, had readily available mpox testing and were familiar with mpox management.

While some previous studies found no difference in clinical presentation based on HIV status,9 14 CD4 count may be more reflective of risk than HIV status itself.12 Unfortunately, we only had self-reported HIV status, and no CD4 count or other markers of immunocompromise available to us. It should be noted that the vast majority of people living with diagnosed HIV in England are on ART, well managed and are virally suppressed, (this group may also reflect different demographics to those few who are not)30 but why there is a lower risk of hospital admission in this group is unclear. If there are true differences in the behaviours and risk of hospital admission among individuals based on HIV and PrEP status, it is important to quantify such differences. Furthermore, understanding potential reasons for differences could impact on planning and preparedness for similar outbreaks in the future.

Given the observational nature of this study, this finding on PrEP usage could be a genuine effect or a false finding due to unadjusted residual confounding from missing data or potential limited representativeness and bias of the small sample (19.4% of all cases) who responded to the questionnaire. It is possible that the characteristics of patients who responded to the questionnaire are different from those who did not. We also relied on self-reported symptoms reported at a single point in time and medical history including HIV and PrEP status. By using self-reported symptoms we could also not make an objective measure of the severity of symptoms. Furthermore, we did not have access to clinical records and comorbidity records or laboratory markers and therefore, did not account for these potential confounders.

Limitations of this study must be noted. First, we were unable to link 11% of patients to their hospital records due to missing key identifiers. This may result in an underestimate of the numbers of ED attendances and inpatient admissions presented. Second, we applied pragmatic criteria based on time interval between date of positive laboratory test and hospital episodes. A patient admitted outside the time window would have been excluded unless they had an mpox ICD code in the SUS record. It is possible that some hospital usage in the specified time window were unrelated to mpox which would have overestimated the attendance and admission rate due to mpox. As we used time windows as the main method for counting mpox-related hospital episodes, this method could overestimate the numbers for both ED attendances and inpatient admissions. Third, we included only patients with laboratory-confirmed mpox (although we presume most cases in England would have had a positive test) and hence this paper does not present hospital episodes relating to patients without a laboratory test. Finally, the low response rate for questionnaires may have introduced bias, affecting the findings of the multivariable analysis.

Despite these limitations, this comprehensive analysis of data relating to all adult male cases with confirmed mpox infection is a realistic estimate of hospital usage of the 2022 outbreak in England. So, while our analysis of the association of PrEP status on hospital usage for mpox requires further exploration, our findings on the impact on healthcare services can be considered robust.

Conclusions

The 2022 mpox outbreak had a considerable impact on health services in England through hospital admissions, ED attendances and bed days. It also impacted on the work of sexual health team diagnosing, testing and vaccinating patients. There was some variation across regions. London saw the greatest impact and accounted for most cases and hospital usage. Commissioners of sexual and healthcare services should review plans for healthcare provision for similar STI or novel emerging infection outbreaks among GBMSM or naïve populations in the future.

Compared with people living with HIV, we found men without a diagnosis of HIV who did and did not take PrEP were at higher risk of inpatient admission. Further studies are needed to confirm and identify reasons for the higher likelihood of hospital admission for GBMSM without HIV infection.

Data availability statement

Data are available upon reasonable request. Due to the risk of deductive disclosure, we are unable to share the raw data. UKHSA operates a robust governance process for access to protected data, where it is lawful, ethical and safe to do so. Organisations looking to access UKHSA data for public health purposes can contact DataAccess@ukhsa.gov.uk to request an application pack. Further information can be found on (https://www.gov.uk/government/publications/accessing-ukhsa-protected-data/accessing-ukhsa-protected-data).

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

Ethical approval for this study was obtained from UKHSA (REGG number NR0359).

Acknowledgments

We gratefully acknowledge the support of Obaghe Edeghere, Alicia Barrasa, Charlotte Anderson, Harshana Liyanage, Lorenzo Cattarino and Paula Blomquist in conducting this study.