The Cost of Adrenal Insufficiency in England—Analysis of NHS HES Data

ABSTRACT

Background

Despite steroid replacement therapy, patients with adrenal insufficiency (AI) experience excessive infections and related hospital admissions. However, data examining the evolution of admissions, healthcare resources utilisation and cost burden is sparce.

Design

Analysis of National Hospital Episode Statistics (HES) data set which contains details of all admissions and outpatient appointments at NHS hospitals in England.

Methods and Measurements

Spells spanning financial years 2018/19 to 2022/23, focusing on HES codes E27.1 (Primary Adrenocortical Insufficiency; PAI) [n = 57,125], E27.2 (Addisonian Crisis; AC) [n = 12,640] and E27.4 (Other and Unspecified AI; UAI) [n = 79,965] were analysed for admissions, main diagnosis, bed‐days, costs, follow‐up, and readmissions.

Results

Over the study period, admissions for AC remained stable; admissions involving UAI increased, whereas PAI admissions reduced transiently during COVID‐19. Mean length of stay for AC increased from 5.1 to 6.8 days (34%). Patients with primary pneumonia and AI had longer hospital stays than those without AI and were more likely to require critical care. Mean cost per hospital stay increased, rising 25% for AC since 2019/20, reaching £2959 per stay. 10% of patients had >1 non‐elective readmission within 12 months. Endocrinologist follow‐up was lower than expected. Centres treating > 225 spells/year reviewed 20−46% of AI patients within 26 weeks of admission, and only 46% with a main diagnosis of AC in 2022/23.

Conclusions

AI admissions have increased since 2018/19. Bed‐days and cost for AC episodes have also risen. Patients with concomitant AI were more likely to have longer stays and be re‐admitted. Endocrinology follow‐up appears surprisingly low despite published guidelines.

1. Introduction

Adrenal insufficiency (AI) is a complex condition that adversely impacts patients' morbidity [1], and quality of life [2]. AI can be characterised as primary, secondary (often pituitary tumour‐related), or tertiary (usually linked to long‐term steroid use) disease. Adrenal Crises (ACs) are an acute life‐threatening complication of AI with a reported mortality of 6% [3]. ACs can occur in all forms of AI but are more common in primary AI. ACs are often precipitated by an infection or the relapse of a pre‐existing chronic disease. However, ACs are potentially preventable, and outcomes are improved by early initiation of high‐dose glucocorticoids [4, 5]. Regular follow‐up of patients by an endocrinologist can also reduce the long‐term impact of AI and prevent AC [6].

The 2020 UK guidance for the prevention and emergency management of adult patients with AI [5] recommends referring the patient with AC to an endocrinologist before discharge, for further advice on diagnosis, starting regular oral steroids or tapering steroids back to usual dose, and education regarding ‘sick day rules’. However, no guidance is provided within this guideline regarding further follow‐up after an AC. The 2016 guidelines of the Society for Endocrinology for the emergency management of AC [6] recommend after an AC period regular review of the patient by an endocrinologist, initially monthly, and in the long‐term every 6–12 months.

Although AI can have a significant burden on health services [7] a thorough literature review [8, 9, 10, 11] revealed only limited data. To address this gap, we undertook an analysis of Hospital Episode Statistics (HES) data from the National Health Service (NHS) in England. Our objective was to provide valuable insights into current practices, healthcare resource utilisation, and costs associated with the care of AI patients. By doing so we aim to contribute to supporting improved patient outcomes, better healthcare decision‐making, greater efficiency and ultimately to reduce the burden of AI on patients and health services.

2. Materials and Methods

HES collects data from every NHS hospital on a range of activities related to patient care, including diagnoses, treatments, and procedures performed during each hospital episode. An ‘episode’ is a continuous period of care managed by a single consultant. A ‘spell’ is the patients entire stay in hospital and can include a single or multiple episodes [12]. The data also include patient demographics, clinical classifications and administrative information such as admission and discharge dates. Patients are assigned a unique identifier for each episode, so that their care journey can be tracked across different healthcare providers. Patient data is masked and aggregated to protect privacy. The whole process relies on clinical coders to review patients' notes, usually following discharge from hospital. The information is translated into coded data and entered into the hospital's information system. All diagnoses and procedures must be recorded to create a true and factual reflection of each patient's hospital activity. The primary use of HES data is to enable hospitals to be reimbursed for their activity. It is therefore essential that the information in patient notes is clear and accurate. HES data is re‐used with the permission of NHS England, copyright 2024. This work uses data provided by patients and collected by the NHS as part of their care and support. All rights reserved.

Data were collected for episodes of care spanning financial years 2018/19 to 2022/23 (1 April 2018 and 31 March 2023), focusing on HES codes E27.1 (Primary Adrenocortical Insufficiency; PAI) [n = 57,125], E27.2 (Addisonian/Adrenal Crisis; AC) [n = 12,640] and E27.4 (Other and unspecified AI, UAI) [n = 79,965].

We analysed the data to understand; the number of patients and frequency of admissions; the main diagnosis when AI was recorded; sub‐groups of AI diagnoses when AI was recorded as the main diagnosis; the frequency of follow‐up by an endocrinologist; the frequency of readmission within a year of the first recorded admission for AI; the average length and cost of stay; the impact of concomitant AI diagnosis on average length of stay; the frequency of readmission for pneumonia with and without concomitant AI; and the proportion of patients admitted to Intensive Care (ICU) with and without AI but with the same main diagnosis.

3. Results

3.1. Number of Patients Admitted to Hospital With a Main or Secondary Diagnosis of AI

Admitted patients for all three codes (PAI, AC, UAI) fell during the COVID year 2020/21 compared with the preceding year. However, the number of patients admitted with AC remained relatively stable (average 1853 distinct patients per year) throughout the study period. Patients admitted for PAI reduced by 14.3% during the COVID‐19 pandemic (5135 distinct patients in 2020/21 vs. 6175 in 2019/20), thereafter returning to pre‐COVID rates (6490 patients in 2022/23). Patients admitted for Other and Unspecified Adrenocortical Insufficiency (UAI) increased steadily over the past 5 years, from 7050 distinct patients in 2018/19 to 10,345 patients in 2022/23 (Figure 1).

Figure 1.

Number of adrenal insufficiency‐patients admitted by diagnosis and financial year. The total number of patients admitted to hospital with a main or secondary diagnosis of Primary Adrenocortical Insufficiency, Addisonian Crisis or Other and Unspecified Adrenocortical Insufficiency in the financial years 2018/19 to 2022/23.

3.2. Average Number of Spells Per Patient

Individual patients coded as PAI or UAI had a similar number of spells (admissions to hospital) relating to these conditions per year. This rate of admission has been relatively constant over the past 5 years (around 1.9 per patient per year). Patients admitted with AC had a consistently lower level of spells per year, with the latest financial year 2022/23 seeing fewer spells per patient than in previous years (1.31 spells per patient per year compared to mean 1.37 spells per patient over the analysis period (p < 0.05)).

3.3. Main Diagnosis When AI Is Recorded

More than 500 unique main diagnosis codes were recorded for patients admitted with AI during the analysis period. The majority of main diagnoses fell into two main categories:

• Infections & viruses, or the body's reaction to these.

• Exacerbations of chronic conditions (respiratory conditions, chronic kidney disease [CKD] or cardiovascular disease [CVD]).

The most frequently documented main diagnoses codes for patients presenting with AI were J18 (Pneumonia, organism unspecified) and A41 (Other Sepsis) (Figure 2 ), with the exception of the period 2020/21 when COVID‐19 was the most common main diagnosis. COVID‐19 being recorded in 5% of patients presenting with AI during this period and reducing to 3% in 2022/23.

Figure 2.

Most common main diagnoses in 2022/23 by episodes. The most commonly documented main diagnoses recorded for patients admitted with AI during the financial year 2022/23.

3.4. Patients Admitted With a Diagnosis of AI Only

42.1% of recorded AC episodes were documented as the main diagnosis, with only 1.2% having no additional coded conditions on the visit record. Of patients presenting with PAI, only 10.9% had it recorded as the main diagnosis, with 3.3% having no additional coded conditions. The largest number of patients with a main diagnosis of AI were coded as UAI, with 17.8% of episodes presenting as the main diagnosis and the highest likelihood of admission with only this diagnosis recorded at 3.9% in 2018/19, increasing to 5% in 2022/23.

3.5. Frequency of Follow Up Appointments With an Endocrinologist

The follow‐up of patients by Specialist Endocrinologists (of all grades) within 26 weeks of admission to hospital with AI was low in almost all hospitals. In the most recent analysis period (2022/23), at major centres where over 225 spells of AI are treated annually, only between 20% and 46% of AI (primary or secondary) patients had an outpatient appointment booked with an Endocrinologist within 26 weeks of admission This includes appointments which were subsequently cancelled, not attended or rebooked. The centre with the highest rate of Endocrinology follow‐up (with at least 100 spells of AI a year) was recorded at 64% in 2022/23. Nationally, follow up rates of all AI have been relatively consistent, with 34% of spells followed up in 2019/20, dropping to 31% in 2020/21 and recovering to 34% in 2022/23 (p > 0.05).

Only 44% of patients who were admitted with AC as a main diagnosis in 2022/23 were followed up with an Endocrinology outpatient appointment within 26 weeks. This figure has changed little since 2019/20, peaking at 48% in 2021/22 (p > 0.05).

It is worth noting that the NHS provides care to the majority of people in the UK. The private sector does not provide acute care, and accounts for only a minority of patient care in the country and this is therefore unlikely to have had a significant impact on the data.

3.6. Frequency of Admission

The frequency of admission for the whole patient cohort was analysed over the full 5 years of data. The first non‐elective admission within the first 4 years of data was identified for each patient, and then the analysis counted the number of times they were admitted non‐electively within 12 months, with a relevant AI diagnosis. The majority of patients (51.8%) with a main diagnosis of AI had no further admissions within a year of their initial admission and 27.3% were admitted once within the 12 months after their initial admission. Of all patients with a diagnosis of AI, less than 1% were seen more than 8 times within the 12 months of their initial admission. However, a small percentage of patients with AI recorded as a main condition were admitted up to 21 times within 12 months.

3.7. Average Length of Stay and Cost Per Spell

Average length of stay for patients with PAI was 4.0 days in the analysis year 2019/20. This increased through the COVID‐19 period to 4.3 days, but then reduced year on year reaching 4.1 days in 2022/23 (4% higher than in 2019/20—p > 0.05). The length of stay for patients recorded with AC increased continuously between 2019/20 and 2022/23, with an acceleration in the latest year. The most recent figure of an average duration of 6.8 days in hospital is 34% higher than that seen in 2019/20 for these patients (p < 0.05). Length of stay for UAI increased during the COVID pandemic and then subsequently reduced yet remained greater than the baseline; in 2022/23 it was 4.8 days, which is 7.5% higher than in 2019/20 (p < 0.05) (Figure 3).

Figure 3.

Mean length of stay as inpatient (days) and mean cost per spell by diagnosis and financial year.

The average cost per spell in 2022/23 was £2218 for PAI, £2983 for AC; and £2399 for UAI. This has increased for all three diagnoses over the analysis period, with the most marked increase seen in those patients recorded with AC which rose by 25% since 2019/20 (11% in real terms after accounting for general health inflation https://www.ons.gov.uk/economy/inflationandpriceindices/timeseries/d7bz/mm23 accessed October 2024), when the cost was £2470 (Figure 3). (Comparable cost data was not available for the 2018/19 year).

3.8. Impact of Concomitant AI Diagnosis on Length of Stay and ICU Admissions

The data were analysed to establish the extent to which AI impacted the length of stay of patients admitted for other main diagnoses. When AI was recorded, the most common main diagnosis was pneumonia. We therefore used pneumonia as the defined disease to determine if AI increased the duration of admission of the primary disease, and to compare the average duration of stay of pneumonia patients with and without concomitant AI recorded.

In 2021/22 and 2022/23, patients who had a diagnosis of both pneumonia and a concomitant AI diagnosis (n = 923 spells in 2021/22 and n = 1091 in 2022/23) had an average duration of stay between 0.6 and 1.04 days longer than those with no AI diagnosis (n = 212,852 spells in 2021/22 and n = 242,033 in 2022/23; Table 1). Additionally, the presence of AI increased the likelihood of critical care being required, with this being most marked in 2022/23, where the requirement rate increased from 2.6% without AI to 3.7% when AI was additionally present (p < 0.05). In 2022/23, patients with pneumonia requiring critical care had longer duration of stay in critical care when AI was present (17.4 days on average) compared to those without AI (14.5 days on average; Table 1, [p < 0.1]). A large variation in mean ICU days per patient with pneumonia without concomitant AI between 2021/22 and 2022/23 was noted, but the reason for this difference remains unclear.

Table 1.

Comparison of main diagnosis of pneumonia with and without concomitant AI.

	Pneumonia without adrenal insufficiency		Pneumonia with adrenal insufficiency
Year	2021/22	2022/23	2021/22	2022/23
Total days	2,091,277	2,432,887	9,630	12,096
Spells	212,852	242,033	923	1091
Mean length of stay	9.83	10.05	10.43	11.09
Main pneumonia emergency spells	206,761	233,998	906	1073
Main pneumonia emergency spells with ICU	5390	6051	29	40
% spells with ICU	2.6%	2.6%	3.2%	3.7%
ICU days	210,700	87,439	480	696
Mean ICU days per spell	39.1	14.5	16.6	17.4
Readmission rate within 30 days	5.4%	4.6%	8.5%	6.9%

Note: Average length of stay of spells where patients admitted with AI in financial years 2021/22 and 2022/23 comparing those with and without pneumonia. The proportion of emergency spells with a diagnosis of pneumonia requiring critical care, comparing those with and without concomitant AI. Comparison of the percentage of spells with a diagnosis of pneumonia who were readmitted (non‐elective) to hospital with or without concomitant AI.

3.9. Frequency of Readmission for Main Condition

Continuing to use pneumonia as a main condition, we compared two groups: one where AI was coded alongside pneumonia, and another where pneumonia was coded in the absence of AI. Our goal was to determine the percentage of patients who experienced a non‐elective readmission for pneumonia within 30 days after their initial treatment. For each year analysed, those patients with AI in their original spell were more likely to be readmitted subsequently. For the period 2022/23 4.6% were readmitted for pneumonia without AI, compared to 6.9% for pneumonia with AI (p < 0.05). However the proportion of spells resulting in a readmission fell in both groups over the study period (Table 1).

4. Discussion

This retrospective, observational study of almost 150,000 spells of care over a 5‐year analysis period clearly shows a rising number of hospital admissions, increasing length of stay, and associated costs in patients with a diagnosis of AI, particularly for AC. The number of patients admitted to hospital with AI has increased since 2018/19, However, we found the most concerning finding to be that the follow‐up of these patients within specialist Endocrinology clinics was unexpectedly low.

Whilst the number of patients admitted with AC and Primary AI was stable, albeit with a small reduction for Primary AI admissions during COVID‐19, the steady increase in admissions for ‘Other and Unspecified’ AI over the past 5 years is noteworthy. We believe this could be due to a number of factors. Increased recognition of iatrogenic AI as a result of pharmacological use of glucocorticoids (oral, inhaled, and topical), opioid therapy, and the widespread use of immunotherapy are potential reasons for the rise. However, as code E27.4 relates to ‘Other and Unspecified AI’ the increase may also reflect a lack of information recorded in the patients notes as to the aetiology of the AI, making it difficult for the coder to select a more specific code.

A recent Australian study [13] has shown a large increase of admissions for AI (by 298.0%) and Addisonian Crises (by 1992.9%) in young adult females between the years 2005 and 2021, with a particular increase over the last 4 years. While the authors could not identify clear reasons for this variation, changes to the underlying incidence of AI or increases in the prevalence of factors which lead to the need for hospital treatment among AI patients, or a combination of these, are likely to be responsible.

Over the last few years there has been considerable work to standardise the patient held steroid emergency card across Europe and the UK. The European Society for Endocrinology and several professional bodies within the UK, in addition to NHS England, have supported this initiative, including the Royal College of Physicians and Society for Endocrinology. The card was launched by the NHS in August 2020 (https://www.england.nhs.uk/wp-content/uploads/2020/08/NPSA-Emergency-Steroid-Card-FINAL-2.3.pdf; accessed October 2024). The card includes a management summary for the emergency treatment of AC alongside a link to the Society for Endocrinology emergency management guidelines. This has done much to raise the awareness of AI in England. The Steroid Emergency Card recommends hydrocortisone injection as the first measure in the treatment of AC. Further awareness is raised by UK patient groups such as the Pituitary Foundation (https://www.pituitary.org.uk/information/adrenal-insufficiency/) and the Addison's Disease Self‐Help Group (https://www.addisonsdisease.org.uk/). In parallel, work to educate paramedics and emergency clinicians on the importance of prompt treatment with high‐dose glucocorticoids may have contributed to the stability in admissions for AC, despite an increase in overall numbers of AI cases. As such there could potentially have been a number of patients who were seen and successfully treated for AC by emergency care services. Some of these patients may not have been admitted to hospital and therefore might not be recorded in the HES data set. Indeed, a German study [4] showed that patients with AC benefitting from early injection of hydrocortisone by self‐injection, had significantly lower hospital admission rates than those with a longer treatment delay.

The higher cost per admission for all three AI diagnoses over the analysis period is likely multifactorial, with the increased length of stay being a significant contributor. This in itself is interesting and difficult to explain, though may reflect a trend towards uncomplicated cases being managed as out‐patients after initial assessment. Inflation, along with recent increases in the overall cost of care per se over time will also be contributing to this rise.

Patients with pneumonia spells and a concomitant AI diagnosis had longer average length of stay than those with no AI diagnosis, which could be due to a number of factors. Patients may be held in hospital whilst being weaned off steroids; awaiting specialist endocrine opinion, particularly if they have been admitted to a non‐endocrine ward; or they may be waiting for the availability of endocrine nurses to educate them on managing their condition and medication. These factors can all delay the discharge of the patient from hospital. It is also interesting to note that the cohort with concomitant AI saw a greater increase in length of stay over the period 2022−2023 than the non‐AI cohort. Regardless, these data show unequivocally that patients with AI are becoming more expensive to treat which in turn raises concerns about the potential impact of concomitant AI diagnoses on healthcare resource utilisation.

Another interesting finding was the increase in the proportion of patients requiring critical care with pneumonia when AI was present compared to those who were admitted with pneumonia and no AI. Although this increased during the analysis period, the figure (3.7% in 2022/23) is low compared to a German study [4] where between 20% and 36% of patients experiencing adrenal emergency required critical care. Due to specific HES coding for ICU admissions it was not possible to analyse our data further to identify patients that were admitted to high dependency units rather than ICU.

A significant finding from our analysis is the low rate of follow‐up after an acute episode of AI that required admission. We would have expected the majority of patients to have a follow‐up appointment with a Specialist Endocrinologist within 26 weeks of discharge from hospital. In half of hospitals who manage > 100 AI spells annually, follow‐up rates were below one‐third of spells in the period 2022/23. The reasons for this are unclear, and could be multifactorial, including recent pressures on NHS resources which have led to backlogs in routine and follow‐up care. It may also be that patients with tertiary AI undergo their follow‐up under their initial specialty team. However, there are likely to be different issues in different localities and unfortunately such data cannot be derived from a database of hospital admissions, clinic attendances and illness codes. Follow‐up of AI in the UK has traditionally been undertaken almost exclusively by Diabetes & Endocrinology units. However, with increasing recognition of tertiary causes of endocrinology relating to glucocorticoid therapy and opioids, as well as immunotherapy, it is probable that other non‐endocrine specialist's follow‐up specific patients who develop AI (i.e., Rheumatology and Oncology). The advent of the European steroid card and widespread use by non‐endocrine specialists has likely encouraged this further. Finally, we recommend adherence to published best practice guidelines for the prevention and emergency management of patients with AI [5, 6].

Increasing awareness and education of AI amongst the plethora of healthcare professional who come into contact with these patients is an essential strategy. Hand in hand with this has to be the provision of education to patients, their families, emergency services and secondary care about the treatment of adrenal crises.

5. Limitations

Our findings should be interpreted bearing in mind several limitations in this study.

We used a retrospective observational design, relying on existing data from HES. This can limit the ability to establish causality and may include biases inherent in the data collection process, as per all national databases and registries. Indeed, the accuracy and completeness of the data in HES is dependent on coding, which in turn is dependent on the diagnosis being made by the attending physician and the quality of clinical documentation. It is possible patients could have been coded as having more than one AI diagnosis during a spell and that there has been both under and over diagnosis of adrenal crises. The accuracy of diagnosis cannot be confirmed within this database. Additionally, the coding for admissions to ICU does not provide detailed information about the level of care provided, for example high dependency unit admissions. Our study did not account for potential variables that could influence outcomes, such as patient demographic, or socioeconomic status. As the study focused on HES data from the NHS in England, this may limit generalising the findings to other healthcare systems or regions.

One of the main findings of our study was the low rates of follow‐up appointments with Specialist Endocrinologists following hospital admissions for AI. However, the reasons for these low follow‐up rates cannot be fully explored in the analysis. Understanding the barriers to follow‐up care is essential for informing interventions to improve patient outcomes. Further research is needed to confirm and understand the underlying issues driving the observed trends.

Conflicts of Interest

The authors declare no conflicts of interest.