You are here

  • Home
  • Archive
  • Volume 33, Issue 2
  • Lowering levels of bed occupancy is associated with decreased inhospital mortality and improved performance on the 4-hour target in a UK District General Hospital

Article Text

Article menu
Original article
Lowering levels of bed occupancy is associated with decreased inhospital mortality and improved performance on the 4-hour target in a UK District General Hospital
  1. D G Boden1,
  2. A Agarwal2,
  3. T Hussain2,
  4. S J Martin2,
  5. N Radford3,
  6. M S Riyat1,
  7. K So1,
  8. Y Su4,
  9. A Turvey5,
  10. C I Whale2
  1. 1Emergency Department, Royal Derby Hospital, Derby, UK
  2. 2Division of Medicine, Royal Derby Hospital, Derby, UK
  3. 3Department of Operations, Royal Derby Hospital, Derby, UK
  4. 4Dr Su Statistics, Consulting firm, Kaunakakai, Hawaii, USA
  5. 5Information Services, RDH, Derby, UK
  1. Correspondence to Dr D G Boden, Emergency Department, Royal Derby Hospital, Derby, DE22 3NE, UK; dan.boden1{at}nhs.net

Abstract

Objective To evaluate whether there is an association between an intervention to reduce medical bed occupancy and performance on the 4-hour target and hospital mortality.

Methods This before-and-after study was undertaken in a large UK District General Hospital over a 32 month period. A range of interventions were undertaken to reduce medical bed occupancy within the Trust. Performance on the 4-hour target and hospital mortality (hospital standardised mortality ratio (HSMR), summary hospital-level mortality indicator (SHMI) and crude mortality) were compared before, and after, intervention. Daily data on medical bed occupancy and percentage of patients meeting the 4-hour target was collected from hospital records. Segmented regression analysis of interrupted time-series method was used to estimate the changes in levels and trends in average medical bed occupancy, monthly performance on the target and monthly mortality measures (HSMR, SHMI and crude mortality) that followed the intervention.

Results Mean medical bed occupancy decreased significantly from 93.7% to 90.2% (p=0.02). The trend change in target performance, when comparing preintervention and postintervention, revealed a significant improvement (p=0.019). The intervention was associated with a mean reduction in all markers of mortality (range 4.5–4.8%). SHMI (p=0.02) and crude mortality (p=0.018) showed significant trend changes after intervention.

Conclusions Lowering medical bed occupancy is associated with reduced patient mortality and improved ability of the acute Trust to achieve the 95% 4-hour target. Whole system transformation is required to create lower average medical bed occupancy.

  • death/mortality
  • emergency department
  • quality assurance

https://doi.org/10.1136/emermed-2014-204479

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Key messages

    What is already known on this subject?

    • Known correlation found in previous studies between high bed occupancy and infection risk.

    • Minimal international research has found a correlation between Trust bed occupancy and mortality.

    • No previous studies have looked at bed occupancy and mortality after deliberate intervention to reduce bed occupancy.

    • No UK studies have shown a correlation between bed occupancy and mortality

    What might this study add?

    • Previous studies have found higher mortality as bed occupancy rises but have not studied the effect on mortality following a deliberate intervention to reduce bed occupancy. This retrospective, single centre study shows that, on intervening to reduce bed occupancy, there appears to be an association with a reduction in mortality and improved performance against the 4-hour target.

    Introduction

    Crowding, access block and high workload have been cited as reasons for excess mortality among ED patients.1–7 The problem of crowding in Emergency Departments (EDs) is recognised by professional bodies around the world as a patient safety issue and, in 2014, this was the principle campaign of the UK Royal College of Emergency Medicine.

    Past research has focused on the outcome of ED patients, but less is known about the effect of access block, capacity and performance for the wider group of patients in a hospital. There is increasing evidence of a correlation between Trust bed occupancy and rates of infection, particularly Clostridium difficile,8 which may lead to poor patient outcomes, but there is little published evidence of a correlation between bed occupancy and mortality. A small number of international studies have been published.9–11 One Danish study showed a 9% increase in mortality rates for inhospital mortality and 30-day mortality when high bed occupancy periods were compared with low bed occupancy periods.

    In UK practice the bulk of patients admitted through the ED require acute medical admission. Access block frequently occurs for patients awaiting admission under the general medical teams with a widespread belief that delays and long trolley waits lead to poor departmental performance, poor care and potential harm to patients.

    UK emergency departments are required to report the proportion of patients attending who are seen, treated, admitted or discharged within 4 h or arrival against a national standard of 95%.

    In June 2013, as a result of increasing workload pressures, Derby Teaching Hospitals NHS Foundation Trust introduced a 90% medicine bed occupancy target. A number of interventions were undertaken across the patient journey to facilitate this. These included daily Consultant ward rounds on medical wards, CCG-commissioning of additional community beds and planned utilisation of traditional surgical bed base for medical patients. This permitted a natural experiment to see the effect of this reduction in bed occupancy on 4-h target performance and hospital mortality.

    The aim of this study is to evaluate whether there is an association between an intervention to reduce medical bed occupancy and 4-h target performance and hospital mortality in a UK District General Hospital.

    Methods

    Study design and setting

    We conducted an uncontrolled before-and-after intervention study at Derby Teaching Hospitals NHS Foundation Trust, a large District General Hospital seeing over 140 000 non-elective patients per year. Mortality data were gathered and analysed for the period January 2010–October 2014.

    Definitions and patient outcomes

    We defined bed occupancy as the number of occupied medical beds as a proportion of the total bed base at midnight. Outliers (medical patients on non-medical wards) were included in the numerator. It is therefore possible to have a derived medical bed occupancy of greater than 100%. Bed occupancy data was gathered from January 2012 to October 2014.

    The 4-hour target performance was defined as above. We determined performance against the 4-hour access target on a weekly basis from local data collection.

    Patient outcomes were determined using hospital standardised mortality ratio (HSMR), summary hospital-level mortality indicator (SHMI) and crude mortality. These have been defined and obtained as documented below:

    Hospital standardised mortality ratio

    The HSMR is a ratio of the observed number of inhospital deaths at the end of a continuous inpatient spell to the expected number of inhospital deaths (multiplied by 100) for 56 specific clinical classification system groups.

    Monthly figures obtained from Dr Foster intelligence.12

    Summary hospital-level mortality indicator

    The ratio between the actual number of patients who die following hospitalisation at the Trust and the number that would be expected to die on the basis of average England figures, given the characteristics of the patients treated there. It covers all deaths reported of patients who were admitted to non-specialist acute Trusts in England and either die while in hospital or within 30 days of discharge.

    Monthly figures obtained from internal Trust data.

    Monthly crude mortality

    The number of deaths that occur in a hospital in any given time period compared with the number of patients admitted for care in that hospital for the same time period. The crude mortality rate has then been set as the number of deaths for every 100 patients admitted.

    Monthly figures obtained from internal Trust data.

    Statistical analysis

    Segmented regression analysis of interrupted time-series method13 ,14 was used to estimate the changes in levels and trends in average medicine bed occupancy, monthly 95% four hour target performance (4HTP) and monthly mortality measures (HSMR, SHMI and crude mortality) that followed the intervention.

    The model used was Yt=β0+β1*T1t+β2*It+β3*T2t+et

    Coefficient β0 estimated the monthly value of the outcome variable at time 0 (just before the beginning of the observation period, January 2012); β1 estimated the baseline slope parameter representing change in the outcome variable of interest that occurred every month before the intervention; β2 was change in the outcome variable of interest immediately after the intervention (intercept changes); β3 estimated monthly change in outcome variable of interest compared with trend before the intervention (slope changes).

    Note that Yt is the outcome variable in month t; T1t is a continuous variable indicating time in months at time t from the start of the observation period (T1=1, 2, …, 34); It is an indicator variable equal to 0 before the intervention and equal to 1 after the intervention; T2t represents time after intervention, equal to 0 before the intervention and equal to the number of months after the intervention; the error term et represents the random error not explained by the model, consisting of normally distributed random error and an error at time t that may be correlated to errors at preceding time points.

    The generalised Durbin-Watson statistic15 was calculated to test for the serial autocorrelation of the error terms in the regression models. When autocorrelation existed, the stepwise autoregression process was conducted using the Yule-Walker method to correct for autocorrelation. Seasonality was accounted by the inclusion of autocorrelation errors.16 Initial autoregressive parameters were set to 13 as the data were collected at the monthly level to account for seasonality. The ith order autoregressive error was retained in the model if the term contributed significantly to the fit of the model (p<0.05). Using the stepwise autoregression process, the following autoregressive errors were retained: medical bed occupancy (no autoregressive error retained), 4 h target performance (no autoregressive error retained), HSMR (a 3rd order autoregressive error retained), SHMI (a 7th order autoregressive error retained), crude mortality (a 7th order and an 11th order autoregressive errors retained). The statistical package SAS V.9.317 was used for all analyses. A p value less than 0.05 was considered statistically significant.

    Results

    Preintervention and postintervention data from January 2012 to October 2014 is summarised in table 1 below.

    View this table:
    Table 1

    Monthly preintervention and postintervention data (January 2012–October 2014)

    Total Trust attendances during this period were 210 510. Attendances increased, with a monthly mean of 11 695 attendances before the intervention and 12 003 following the intervention. Non-elective admissions also increased with a mean of 2986 preintervention compared with 3263 postintervention.

    Medical bed occupancy

    Following the planned intervention in July 2013 mean medical bed occupancy decreased from a preintervention mean of 93.7% to a postintervention mean of 90.2%, (see table 1).

    A statistically significant level change (p=0.02) in bed occupancy was noted. However, due to a slight rise in occupancy levels in September and October 2014, there was a resultant non-significant trend change after intervention (p=0.29) (figure 1, table 2).

    View this table:
    Table 2

    Interrupted time-series regression analysis of average medical bed occupancy

    Figure 1
    Figure 1

    Average monthly medical bed occupancy (January 2012–October 2014).

    Four-hour target performance

    Four-hour target performance was compared with midnight bed occupancy in medicine. Following intervention performance on the target improved with the Trust achieving a 95% or greater performance in 37/72 (51.4%) weeks as compared with the preintervention period where the standard was achieved in 24/72 (33.3%) weeks (see online supplementary appendix 1)

    The trend change in target performance when comparing preintervention and postintervention revealed a statistically significant improvement (p=0.019) (figure 2, table 3).

    View this table:
    Table 3

    Interrupted time-series regression analysis of monthly 95% target performance

    Figure 2
    Figure 2

    Monthly 95% target performance (January 2012–October 2014).

    Medical bed occupancy and mortality

    With respect to the relationship between medical bed occupancy and mortality before and after intervention, table 1 shows data on monthly average medical bed occupancy, HSMR, SHMI and crude mortality with accompanying relevant CIs. Total Trust attendances, number of deaths and non-elective admissions are also included to illustrate the total deaths in relation to number of admissions and attendances.

    The intervention resulted in mean reductions in all markers of mortality (range 4.5–4.8%). SHMI (p=0.02) and crude mortality (p=0.018) resulted in significant trend changes after intervention (figures 3⇓–5, tables 4⇓–6).

    View this table:
    Table 4

    Interrupted time-series regression analysis of monthly HSMR

    View this table:
    Table 5

    Interrupted time-series regression analysis of monthly SHMI

    View this table:
    Table 6

    Interrupted time-series regression analysis of crude mortality

    Figure 3
    Figure 3

    Monthly hospital standardised mortality ratio (HSMR) (January 2012–October 2014).

    Figure 4
    Figure 4

    Monthly summary hospital-level mortality indicator (SHMI) (January 2012–October 2014).

    Figure 5
    Figure 5

    Monthly crude mortality (January 2012–October 2014).

    Discussion

    After the introduction of relevant interventions to reduce bed occupancy in medicine, our principle findings are statistically significant differences in medical bed occupancy (level change), 4 h target performance and SHMI and crude mortality. Where this study differs from previously published research is that the association of bed occupancy on mortality has been studied after specific interventions were undertaken to reduce bed occupancy. This has not previously been studied and is potentially reproducible by other hospitals and Trusts.

    The data in this before-and-after study shows an association between these factors but we must be cautious in the interpretation of such data as association does not equate to causality.

    First, the limitations of any before-and-after study design must be noted. Undertaking such a study type runs the risk of conclusions being made that actually result from secular/temporal trends, of regression to the mean, of potential for influence from the Hawthorne effect and challenges with regards to sustainability and generalisability.

    In addition a number of confounding factors may influence the findings and these are detailed in table 7.

    View this table:
    Table 7

    Confounders and their potential influence

    We feel it is important to highlight these other variables that may have contributed to the reduction in mortality that we have found in this study. This is because, at present, it remains difficult to disentangle these (and other) confounders from the results. By doing this we hope to open debate as to the potential role that all of these may have played. This, in turn, should increase our knowledge and understanding in a vitally important area of Emergency Medicine.

    Even accounting for the challenges posed by confounders, we believe that the negative patient and system outcomes associated with high bed occupancy pass face validity and deserve further consideration and research.

    In 2014 the Royal College of Emergency Medicine published a paper titled ‘Crowding in the Emergency Department’.19 Within this document they state that an Emergency Department is crowded if “ambulances cannot offload, there are long delays for high acuity patients to see a doctor, there are high rates of patients with a ‘Left Without Being Seen’ Code, there are more trolley patients in the ED than there are cubicle spaces, or if patients are waiting more than 2 hours for an inpatient bed after a decision to admit has been made”.

    Studies in a number of different countries demonstrate increasing evidence of the importance of ED crowding and its detriment to patient care including:

    • Evidence of an association between crowding and mortality.3–5

    • Reduction in the quality of patient care received.20

    • Increased length of stay for non-elective admissions.21

    • Cancellation/postponement of hospital elective activity22

    All these factors are likely to reduce patient experience and outcome. This paper provides statistical evidence to support the beliefs and experiences of UK and international physicians. The tagline that ‘crowding kills’ is emotive, but important. If our access systems fail and patient harm results then we have a responsibility to monitor and report the data that demonstrates potential causes and associations such that the profession, healthcare systems and patients can explore, understand and improve care.

    This study represents the experience of one Trust over a relatively short time period. We suggest that further research should be conducted in other centres of different size and location to ascertain whether these findings are generalisable to a UK, or even international, population.

    Conclusion

    Lowering medical bed occupancy is associated with reduced patient mortality and improved ability of the acute Trust to achieve the 95% 4 h target. Whole system transformation is required to create lower average medical bed occupancy.

    References

    1. ↵
      1. Forero R,
      2. McCarthy S,
      3. Hillman K
      . Access block and Emergency Department overcrowding. Crit Care 2011;15:216. doi:10.1186/cc9998
      OpenUrlCrossRefPubMed
    2. ↵
      1. Hoot NR,
      2. Aronsky D
      . Systematic review of emergency department crowding: causes, effects and solutions. Ann Emerg Med 2008;52:126–36. doi:10.1016/j.annemergmed.2008.03.014
      OpenUrlCrossRefPubMedWeb of Science
    3. ↵
      1. Richardson DB
      . Increase in patient mortality at 10 days associated with Emergency Department overcrowding. Med J Aust 2006;184:213–16.
      OpenUrlPubMedWeb of Science
    4. ↵
      1. Morris ZS,
      2. Boyle A,
      3. Beniuk K, et al
      . Emergency Department crowding: towards an agenda for evidence-based intervention. Emerg Med J 2012;29:460–6. doi:10.1136/emj.2010.107078
      OpenUrlAbstract/FREE Full Text
    5. ↵
      1. Viccellio A,
      2. Santora C,
      3. Singer AJ, et al
      . The association between transfer of Emergency Department boarders to inpatient hallways and mortality: a 4-year experience. Ann Emerg Med 2009;54:487–91. doi:10.1016/j.annemergmed.2009.03.005
      OpenUrlCrossRefPubMedWeb of Science
    6. ↵
      1. Boyle A,
      2. Beniuk K,
      3. Higginson I, et al
      . Emergency Department Crowding: time for interventions and policy evaluations. Emerg Med Int 2012;2012:1–8. doi:10.1155/2012/838610
      OpenUrl
    7. ↵
      1. Reeder TJ,
      2. Garrison HG
      . When the safety net is unsafe: real-time assessment of the overcrowded Emergency Department. Acad Emerg Med 2001;8:1070–4. doi:10.1111/j.1553-2712.2001.tb01117.x
      OpenUrlCrossRefPubMedWeb of Science
    8. ↵
      1. Ahwoy LC,
      2. Lambert PC,
      3. Jenkins DR, et al
      . Bed occupancy rates and hospital-acquired Clostridium difficile infection: a cohort study. Infect Control Hosp Epidemio 2013;34:1062–9. doi:10.1086/673156
      OpenUrlCrossRef
    9. ↵
      1. Madsen F,
      2. Ladelund S,
      3. Linneberg A
      . High levels of bed occupancy associated with increased inpatient and thirty-day hospital mortality in Denmark. Health Aff (Millwood) 2014;33:1236–44. doi:10.1377/hlthaff.2013.1303
      OpenUrlAbstract/FREE Full Text
    10. ↵
      1. Schilling PL,
      2. Campbell DA,
      3. Englesbe MJ, et al
      . A Comparison of in-hospital mortality risk conferred by high hospital occupancy, differences in nurse staffing levels, weekend admission and seasonal influenza. Med Care 2010;48:224–32. doi:10.1097/MLR.0b013e3181c162c0
      OpenUrlCrossRefPubMedWeb of Science
    11. ↵
      1. Sprivulis PC,
      2. Da Silva JA,
      3. Jacobs IG, et al
      . The association between hospital overcrowding and mortality among patients admitted via Western Australian emergency departments. Med J Aust 2006;184:208–12.
      OpenUrlPubMedWeb of Science
    12. ↵
    13. ↵
      1. Wagner A,
      2. Soumerai S,
      3. Zhang F, et al
      . Segmented regression analysis of interrupted time series studies in medication use research. J Clin Pharm Ther 2002;27:299–309. doi:10.1046/j.1365-2710.2002.00430.x
      OpenUrlCrossRefPubMedWeb of Science
    14. ↵
      1. Penfold R,
      2. Zhang F
      . Use of interrupted time series analysis in evaluating health care quality improvements. Acad Pediatr 2013;13:S38–44. doi:10.1016/j.acap.2013.08.002
      OpenUrlCrossRefPubMed
    15. ↵
      1. Ansley C,
      2. Kohn R,
      3. Shively T
      . Computing p-Values for the Generalized Durbin-Watson and Other Invariant Test Statistics. J Econometr 1992;54: 277–300. doi:10.1016/0304-4076(92)90109-5
      OpenUrlCrossRef
    16. ↵
      1. Brocklebank J,
      2. Dickey D
      . SAS for Forecasting Time Series. Cary, NC: SAS Institute Inc., 2003.
    17. ↵
      SAS Institute Inc. SAS/STAT 9.3 user's guide. Cary, NC: SAS Institute Inc., 2012.
      1. Jacques RM,
      2. Fotheringham J,
      3. Campbell MJ, et al
      . Did hospital mortality in England change from 2005 to 2010? A retrospective cohort analysis. BMC Health Serv Res 2013;13:216. doi:10.1186/1472-6963-13-216
      OpenUrlCrossRefPubMed
    19. ↵
      College of Emergency Medicine. Crowding in Emergency Departments.secure.collemergencymed.ac.uk/code/document.asp?ID=6296 Rev June 2014.
    20. ↵
      1. Tekwani K,
      2. Kerem Y,
      3. Mistry C, et al
      . Emergency Department overcrowding is associated with reduced satisfaction scores in patients discharged from the Emergency Department. Western J Emerg Med 2013;14:11–15. doi:10.5811/westjem.2011.11.11456
      OpenUrlCrossRef
    21. ↵
      1. Krochmal P,
      2. Riley TA
      . Increased healthcare costs associated with ED overcrowding. Am J Emerg Med 1994;12:265–6. doi:10.1016/0735-6757(94)90135-X
      OpenUrlCrossRefPubMedWeb of Science
    22. ↵
      1. Mayor S
      . Hospitals take short term measures to meet targets. BMJ 2003;326:1054.
      OpenUrlFREE Full Text

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    • Press release

      Files in this Data Supplement:

    Footnotes

    • Contributors The concept of the potential association between medicine bed occupancy and mortality and the 95% target was conceived by DGB (Emergency Medicine Consultant). Other contributions have been made, in alphabetical order, by AA (Workstream Four—facilitating ward discharge in the Acute Trust), TH (Workstream Two Lead—Medical Assessment Unit), SJM (Divisional Director for Medicine and winter planning team member), NR (Head of Operations—coordinated winter planning), MSR (Emergency Medicine Consultant—Cardiac arrest data), KS (Emergency Medicine Trainee—Cardiac arrest data), YS (Statistician), AT (Trust Mortality data) and CIW (Respiratory Consultant and Workstream Three Lead—coordinating uniformed working on medical base wards). DGB is the nominated guarantor of the article.

    • Competing interests None declared.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    Linked Articles