1 June 2018

Jessica Morris

Research Analyst
Nuffield Trust

Introduction

Emergency readmissions – where patients are readmitted to hospital in an emergency within 30 days of discharge – are frequently used as a measure of the quality of care provided by a health care system. They are also used as an indicator for when poor patient outcomes could potentially have been avoided. For a long time emergency readmissions have been a key issue for policy-makers, as it is thought that a reduction in readmissions can simultaneously improve health care quality and reduce overall costs.

Following a decade which saw a significant increase in emergency readmissions in England, in 2011/12 the Department of Health introduced a financial incentive known as the 30-day emergency readmission rule [1]. This has meant that hospitals are no longer reimbursed for certain types of emergency readmissions above locally agreed thresholds.

NHS Digital have not published data on emergency readmissions since December 2013, as they are delaying indicator updates while they review the methodology [2].
In October 2017, Healthwatch England published a briefing which evaluated 30-day emergency readmissions for 72 trusts that responded in full to their information request, including a day-by-day breakdown of numbers of readmissions in 2013/14 compared to 2016/17 [3]. Our analysis will expand on this report to examine trends in emergency readmissions for all acute hospital trusts in England using Hospital Episode Statistics (HES) data (see Appendix).

Our aims are:

  • to add to the data on emergency readmissions that is currently available in the public domain by analysing data for all NHS hospital trusts in England
  • to focus on trends in emergency readmissions that were potentially preventable as this may highlight areas for quality improvement.

Our approach

High numbers of emergency readmissions do not simply signal hospitals sending patients home too quickly. Some emergency readmissions may result from potentially avoidable adverse events, but others may be due to unrelated or unforeseen causes of admission.

Some may relate to changes in the way that hospitals run services – for example through the increased use of frailty and ambulatory care units. And others might be a consequence of our ageing population and the increase in the number of people living with multiple chronic conditions which often require hospital care.

A recent longitudinal study which calculated risk-adjusted 30-day readmission rates (which take account of changes in patient characteristics over time) found that rates remained relatively stable, but that there were large variations in trends across clinical areas, with some experiencing noticeable increases in readmissions [4].

Therefore, as well as analysing trends in the total number of emergency readmissions in England, we focused on identifying those readmissions that were “potentially preventable”. For this we used combinations of diagnosis and admission codes to indicate where altered care might potentially have prevented readmission. We followed a similar methodology to our paper Classifying emergency 30-day readmissions in England using routine hospital data 2004-2010: what is the scope for reduction? To account for the overall increase in hospital admissions over time we evaluated both numbers and rates of emergency readmissions.

Our findings

Our analysis shows that between 2010/11 and 2016/17 the total number of 30-day emergency readmissions to hospital in England increased by 19.2%, from 1,157,570 to 1,379,790. Meanwhile, the total number of hospital admissions increased by 10.5% over the same time period, from 15,527,166 to 17,164,662. The emergency readmissions rate increased from 7.5% in 2010/11 to 8.0% in 2016/17 (Figure 1).

Figure 1. Emergency readmissions

We examined the day-by-day breakdown of the number of people readmitted within the 30-day period after discharge (Figure 2). Our results show a similar pattern to those produced by Healthwatch England – that the number of emergency readmissions has been increasing year-on-year and that the highest proportion of readmissions occurs one day after discharge, diminishing thereafter. Between 2010/11 and 2016/17, the number of emergency readmissions occurring one day after discharge increased by 24.8%, from 120,239 to 150,060.

It is possible that some emergency readmissions occurring on day zero might be due to hospital transfers and may be a coding issue, rather than people being discharged and readmitted on the same day. The small increases in readmissions that are observed on days 7, 14, 21 and 28 are likely to be due to people attending outpatient appointments a week or two weeks post-discharge who are identified by clinicians as requiring an emergency readmission to hospital.

Figure 2. Number of emergency readmissions

According to our classification of “potentially preventable” emergency readmissions (see Appendix for more information) they increased by 41.3% between 2010/11 and 2016/17, from 130,760 to 184,763 (Figure 3). When accounting for the overall increase in hospital admissions over time, we found that the potentially preventable emergency readmissions rate increased from 0.8% to 1.1% over the same time period.

Our defined group of potentially preventable emergency readmissions accounted for a relatively small proportion of total emergency readmissions; however, we found that this proportion increased over time from 11.3% in 2010/11 to 13.4% in 2016/17. Note that this does not exactly represent the number of preventable readmissions. Our grouping was based on the presence of specific diagnosis codes in the data, but readmissions with these codes are not always preventable: only a case review would find that out. Also, there might be readmissions with other diagnosis codes that were potentially preventable – they are just harder to detect (if at all) in the available data.

Figure 3. Potentially preventable emergency readmissions

Our categorisation of potentially preventable emergency readmissions enabled us to stratify patients by certain diagnoses that we identified as being markers of probable or possible suboptimal care. Note that these are diagnoses which occurred in the first episode of the readmission but did not occur in the index admission. For example, it may be assumed that a patient readmitted to hospital with a surgical or medical complication (such as rupture of an operation wound or an infection following a procedure) may not have received the quality of care that was required.

Figure 4 shows that the 30-day emergency readmissions rates for patients diagnosed in the first episode of readmission with pneumonia, pressure sores or venous thromboembolism increased considerably over time. The emergency readmissions rate for pneumonia increased from 0.26% of admissions in 2010/11 to 0.41% of admissions in 2016/17. The emergency readmissions rate for pressure sores increased from 0.05% to 0.13% and the readmissions rate for venous thromboembolism increased from 0.11% to 0.13% over the same time period.

In terms of absolute numbers, the number of patients that were diagnosed in the first episode of readmission with pneumonia increased from 41,003 in 2010/11 to 70,731 in 2016/17 (a 72.5% increase). The number of patients that were readmitted with a pressure sore increased from 7,787 to 22,448 (a 188.3% increase), and the number of patients that were readmitted with venous thromboembolism increased from 16,890 to 23,006 (a 36.2% increase).

Figure 4. Potentially preventable emergency readmissions

To put these trends into context it is important to look at overall trends in diagnoses for these conditions over time. Changes in coding practices and population changes such as our ageing population and the increase in multimorbidity may have contributed to the trends in disease prevalence on readmission that we observed.

Between 2010/11 and 2016/17, we found that overall diagnoses for pneumonia, pressure sores and venous thromboembolism increased across all hospital episodes by 63.5%, 153.1% and 28.6% respectively. However, the increases in prevalence that we observed for these conditions during the first episode of readmission were greater, increasing by 72.5%, 188.3% and 36.2% respectively.

Implications for quality

Our analysis found that the absolute number and rates of emergency readmissions have increased over time. One possible cause is an older population with more complex conditions. Our analysis did not use risk adjustment (an analytical technique to take account of differences in the characteristics of a population which may impact on the risk of an event).  As such, changes in readmission rates over time may reflect differences in the patient population, with more severely ill and older patients being more likely to be admitted. When taking account of this via risk adjustment, the rate of readmissions is broadly stable so quality of care seems to have been maintained [4].

However, while this aspect of quality has not deteriorated it does not appear to have improved either, so there remains a potential opportunity to target improvement efforts, by focusing on potentially preventable readmissions which have increased at a faster rate than other readmissions.  Notably, we found that the number of people being readmitted to hospital within 30 days of discharge with pneumonia, pressure sores and venous thromboembolism increased at a rate that was greater than other conditions.

In order to fully understand the implications of our findings for practice it is useful to reiterate our definition of what constituted a potentially preventable readmission. These patients were coded as having these conditions during the first episode of readmission but did not have the condition during their index admission. Therefore, we are not providing information about patient groups that are most likely to be readmitted. Rather, we have identified conditions that patients are being readmitted with that probably or possibly could have been prevented if their care was of a higher quality. This may relate to the quality of care during their initial hospital stay, the robustness of discharge planning and also the quality of transitional, community or social care services.

Hospitalisation for community-acquired pneumonia is higher in older people and in people with multiple chronic conditions [5]. Therefore, our ageing population and the increase in multimorbidity over time may have contributed to the increase in the rate of pneumonia readmissions that we observed. However, the increase in pneumonia readmissions was greater than the overall increase in pneumonia diagnoses over time. This indicates that the quality of care may have been compromised, as people are being readmitted with a new diagnosis of pneumonia within a few days of being discharged.

It is widely acknowledged that many pressure sores are preventable, and that when they occur they can be incredibly painful and debilitating for patients. There has been an increased focus on preventing pressure sores in recent years [6], and this has probably improved the recording of pressure sores in patient notes. This improved coding in hospital data may have contributed to the increase in readmission rates for pressure sores. Nevertheless, this increase superseded the overall increase in diagnoses of pressure sores in all hospital episodes. Therefore, it is possible that we have identified an area for improvement in the prevention of pressure sores – either in acute hospital settings or in community and social care providers.

Venous thrombo-embolism (VTE) is one of the leading causes of preventable hospital deaths [7]. Identifying patients who are at risk and the use of thromboprophylaxis can significantly reduce the associated morbidity and mortality. Since 2010, NHS providers have been required to report the proportion of patients who have been risk assessed for VTE [8]. Risk assessments increased rapidly between 2010 and 2012, but have since remained at around 95%. Improved detection and coding as a result of VTE assessments may partly explain the increase in people being readmitted with VTE. Nonetheless, it is possible that the proportion of patients being risk assessed could be improved, and that some patients that ought to have prophylaxis are being missed.

Conclusion

There has been an increase in the absolute number of emergency readmissions in recent years, which highlights the growing and increasingly complex nature of demand faced by the NHS.  Some of these cases may potentially have been prevented: as the National Audit Office has noted  they can act as a “warning indicator” for local providers that they may not be providing the required quality of acute care and discharge planning [9]. We identified some clinical areas that may be of particular concern. Given the national-level policies aimed at reducing readmissions, it is vital that we continue to investigate the data both at a national and local level [10]. This will enable us to target areas for quality improvement and will help us to provide the best level of care for our patients.

Appendix

Emergency readmissions were defined as patients that were readmitted to hospital within 30 days of discharge between 2010/11 and 2016/17.  Readmissions from all NHS trusts were included.  Only ordinary admissions were included, i.e. overnight stays, excluding day cases, regular day or night attenders (for example chemotherapy of dialysis patients), and maternity admissions. Emergency admissions and readmissions were identified using the admission method code.  The analysis is based on admissions rather than patients – so patients can be included more than once.  The denominator for rates was all admissions (not just ordinary admissions), because these all represent an opportunity for a preventable issue to arise requiring readmission. The “potentially preventable” emergency readmissions were categorised using a combination of diagnosis and admission codes to indicate where readmission may have resulted from probable or possible suboptimal care. Note that the diagnosis codes occurred in the first episode of the readmission but not the index admission.
The ICD-10 codes that were used to identify diagnoses in the Hospital Episode Statistics data were:

  • Complications of surgical and medical care, not elsewhere classified (T80-T88)
  • Sequelae of injuries, of poisoning and of other consequences of external causes (T90-T98)
  • Venous thromboembolism (I26.0, I26.9, I63.1, I63.4, I74, I80, I81, I82, T79.0, T79.1)
  • Pneumonia (J12-J18)
  • Pressure sores (L89)
  • Poisoning by drugs, medicaments and biological substances (T36-T50).

References

1. Department of Health. Payment by Results Guidance for 2012-13. London : s.n., 2012. 
2. NHS Digital. NHS Digital Indicator Portal.
3. Healthwatch England. What do the numbers say about emergency readmissions to hospital? 2017.
4. Friebel R, Hauck K, Aylin P, Steventon A. National trends in emergency readmission rates: a longitudinal analysis of administrative data for England between 2006 and 2016. BMJ Open  2018.
5. NICE. Pneumonia: diagnosis and management of community- and hospital-acquired pneumonia in adults.
6. NHS Improvement. Stop the pressure.
7. BMJ Best Practice. VTE prophylaxis. 
8. NHS England. Venous Thromboembolism (VTE) Risk Assessment. 
9. National Audit Office. Reducing emergency admissions. 
10. House of Commons Public Administration and Constitutional Affairs Committee. Follow-up to PHSO report on unsafe discharge from hospital: Government Response to the Committee's Fifth Report of Session 2016-17.

Comments

it would be really useful to understand how many/%age of these readmissions come from care/nursing homes rather than from a person's home. Is that data available?
Martin Heuter (not verified)
(changed )

Thank you. A very interesting report. I have two queries. Can you confirm that the figure of 17,164,662 for "total" hospital admissions in 2016/17 is ALL admissions, not just emergency admissions (which I understood from NHS England stats to be 5,800,183) and includes also elective surgical and day and maternity cases, etc? Second, you emphasise that the potentially preventable readmissions for pneumonia, etc were identified by you from the discharge coding of those readmitted cases and did not apply to the corresponding patients' index admissions. I assume this is because they were not coded in those index admissions. But, I take it that you are not excluding the possibility that those complications in some cases could have existed at the point of the index discharge but not coded, pressure sores being an obvious example? Thank you in anticipation. I am a retired consultant physician and interested in hospital activity.
Dr Stefan D Slater (not verified)
(changed )

Hi Martin, Yes, that's available in HES data via the source of admission field. A related point is that we have previously found that, in a high proportion (30% or so) of admissions with pressure sores, those sores were already present on admission, i.e. not developed during that hospital stay. There's more work to do on prevention in the community, including care homes. Alex
Alex Bottle (not verified)
(changed )

Thank you very much for your comment. It would be very useful to have this sort of data, but unfortunately this isn’t available in Hospital Episode Statistics in any reliable form.
Jessica Morris (not verified)
(changed )

Add new comment