The ageing population of western countries, increasing health demands and crisis of traditional primary care services1 push more and more patients towards overcrowded emergency departments (EDs).2 The risks of inappropriate discharges and unnecessary hospitalisations in this context are substantial. In the last decades “observation medicine”3 has been offered as the solution to balance the limited availability of hospital resources with the need to assess and treat adequately the patients attending the ED. The way in which this practice has evolved has resulted in many local variations as to the naming, acronym and function of dedicated assessment areas. In fact, besides “short observation units”(SOU), “observation wards” or “clinical decision units” (CDU)4^–8 where the length of stay is normally <24 h, there are other examples such as “medical admission units” (MAU) or “medical assessment and planning units” (MAPU)9 10 where the patients can receive a more prolonged assessment and treatment by emergency physicians, internists or multidisciplinary teams. There is reasonable evidence that the chest pain at low or intermediate risk of an acute coronary syndrome can be effectively managed in this way.11^–14 Many other common clinical conditions, however, are suitable for a short period of observation and treatment with a good chance of discharging the patient safely. Asthma,15 16 abdominal traumas,17 atrial fibrillation,18 heart failure,19 syncope,20 head injuries,21 some infections22 and self-harm23 have been managed in short stay dedicated wards. Nevertheless, multipurpose observation wards have been investigated only through case series with historical controls and a large variability in case mix, length of stay, medical specialties involved and setting.24^–26 Little is known about the patients sent home from these units, the consumption of resources and whether any delay in diagnosis and/or treatment occurs. The benefits are not proven but only postulated, as is the effectiveness of current guidelines on their management. A randomised controlled trial could theoretically answer these questions, but its complexity is inherently beyond the scope and capabilities of many non-teaching hospitals such as ours. We have therefore undertaken a “before and after” study to test the reorganisation of our multipurpose SOU, which was operating without rules of admission, clinical policies and strict time limits, adopting the best evidence available, together with the reinforcement of clinical leadership.

METHODS

Setting

Imola Hospital has 293 acute beds for 126 000 people living in a densely populated region of north-eastern Italy. There are approximately 60 000 visits to the ED each year. The ED comprises a main area where the patients with unexpected illness or injury receive the initial care and where the decisions about hospitalisation, observation or discharge are taken by the emergency physicians. The ED has its own emergency ward with 14 beds to which adults can be admitted for continuing assessment and treatment before they are transferred to another hospital ward, health service or discharged. Moreover, there is an SOU near to the ED with three beds used to accommodate patients expected to require a short episode of care, with 24 h access to diagnostic facilities (except exercise tolerance tests and Doppler ultrasound, available only on weekdays). In case of need, one or more beds in the ED can be used (if available) for more patients requiring short observation. Before the intervention, in the morning and afternoon shifts the patients admitted to the ward or the observation unit were managed by the same emergency physician. In the night shift, the emergency physician assigned to the ward/SOU had also to manage the new arrivals in the ED. The theoretical limit of observation was set by the regional health agency at 24 h but, in our hospital, we had neither clear policies of admission to the SOU nor any guideline endorsed to manage these patients. Strict time limits for observation were not respected. The clinical activity was not audited and the adverse events were evaluated on a case-by-case basis.

Intervention

To deal with the temporary closure (due to technical reasons) of one of the two wings of the internal medicine ward, three beds were added to the ED. An adjacent area was arranged to receive up to six patients for observation. The SOU was therefore able to accommodate a total of nine patients. A senior emergency physician (PI) was designated the task of supervising the observation unit.

A set of clear and concise clinical pathways (evidence-based and locally adapted) for most of the conditions amenable to observation (box 1), the objectives and criteria of admission in accordance with the published guidelines4 5 were presented and discussed with the medical and nursing staff. The time limit of 24 h of observation was stressed and the nurses were asked to solicit the physicians to respect it. The nursing staff was increased by one unit for each of the three daily shifts. An emergency physician dedicated exclusively to the patients in the SOU was introduced in the morning shift (08.00–14.00 h) except on holidays. In the night shift (20.00–08.00 h) another physician dedicated to the patients admitted to the ED or SOU was introduced. Moreover, for the patients admitted to the ED with lesser levels of acuity, a policy of fast transfer to the nearest community hospital was adopted.

Analysis of data

The intervention was studied during the period 1 January to 26 March 2007 and compared with the same period in 2006. In both years the follow-up period was 3 months. Age, sex and case mix of the patients admitted for observation were used to compare the baseline features of the two groups. The number of deaths or emergencies during the observation period, number of hospitalisations from the SOU and reattendances after discharge home were the main outcomes. The length of stay, number of consultations and diagnostic procedures in the SOU (ultrasonography in the ED was considered separately) and length of hospital stay were also evaluated. Eighteen groups of syndromes were taken into account. The SOU diagnoses (text and ICD-9 codes) and the final hospital discharge diagnoses (ICD-9 codes) were compared. The eventual discrepancies were resolved by retrieval of the clinical records. The outcomes of patients transferred to the community hospital were also available, being part of the same public health agency (Azienda Sanitaria Locale di Imola).

Data were extracted from the hospital electronic databases and analysed using Excel and Access software programs.

The Student t test for unpaired samples and χ² test (without corrections) were used to compare means and proportions, respectively, with a level of significance set at p<0.05. Direct methods of standardisation were adopted to obtain case mix standardised percentages.

RESULTS

Preliminary findings

There were 9310 attendances at the ED from 1 January to 26 March 2006 compared with 8615 in the same period of 2007; 418 and 762 patients were admitted to the SOU in these two periods, respectively. The baseline characteristics of the studied population are shown in table 1. Age and sex did not differ significantly between the two arms of the study. The proportion of patients admitted to observation rose from 4.5% (418/9310) in 2006 to 8.8% (762/8615) in 2007 (p<0.01).

Hospitalisations

The absolute numbers of hospital admissions (either from the ED or SOU) and discharges from the SOU are shown in fig 1. The mean (SD) age of the patients admitted to hospital from the SOU did not differ between 2006 and 2007 (64.5 (17.8) years vs 65.1 (19.8) years, p = NS). There was also no significant sex difference. Because of the mismatch in case mix, the hospitalisation rates from the SOU were standardised for this variable (see table 2 and fig 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Number of hospital admissions directly from the emergency department (ED), through the short observation unit (SOU) and of cases discharged home from the SOU in 2006 and 2007.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Case mix standardised hospital admissions (×1000 patients admitted to the short observation unit, SOU) before and after the intervention.

The hospitalisations from the SOU increased slightly but not significantly. The most important diagnoses are shown in table 3.

Length of stay and adverse events

After the intervention the length of stay within the SOU was significantly shorter than before (table 4). In 2006 more than 38% of cases exceeded the limit of 24 h compared with less than 4% in 2007 (p<0.01). Adverse events during the stay in the SOU (deaths, emergencies) were very rare. One death occurred just at the beginning of the post-intervention period following a clear violation of the admission rules. The length of hospital stay and deaths after hospital admission did not differ significantly. There were no deaths within the first day of hospital admission from the SOU in either period.

Diagnostic burden

There were fewer consultations and radiological investigations after the intervention (table 5 and fig 3) whereas ultrasonographic examinations in the ED increased significantly. More exercise tolerance tests were requested in 2007 (9 before and 33 after discharge) than in 2006 (1 before and 12 after, p = NS).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Radiological procedures and consultations during the observation period according to the diagnosis.

Follow-up

After the intervention the follow-up showed a significantly lower incidence of reattendances to the ED and of hospitalisations after reattendance (table 6 and figs 4 and 5).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Unplanned reattendances at the emergency department and their outcomes before and after the intervention according to the diagnosis.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Case mix standardised outcomes ×1000 patients discharged from the short observation unit (SOU). All causes and related emergency department (ED) reattendances, hospital admissions after reattendance and repeated SOU within 3 months from discharge.

Seventeen patients in 2006 and 16 patients in 2007 had more than one attendance at the ED. The detailed outcomes of the patients who re-attended the ED after being discharged from the SOU for chest or abdominal pain are shown in fig 6A and B.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6

Outcomes after discharge from the short observation unit (SOU) for (A) chest pain and (B) abdominal pain. Numbers in boxes refer to numbers of cases. GERD, gastro-oesophageal reflux disease.

DISCUSSION

The intervention had a considerable and favourable impact on the overall performance of our multipurpose SOU. It is unlikely that extraneous factors could have contributed to obtain similar results. In particular, there was no evidence that patients observed in 2007 had lesser degrees of disease severity than those in 2006. The two groups were remarkably similar for age and sex. The setting was the same since the comparison was made with patients admitted to the SOU and treated by the same medical team. The mortality and length of stay of patients admitted to hospital from the SOU (considered a “proxy” of disease severity) were similar. The case mix did not differ sufficiently to produce a significant bias. In fact, after standardisation for this variable, the hospitalisation rates did not change significantly and the minor consumption of resources was still confirmed. However, the hospital admission rates were higher than those reported elsewhere,9 26 even if they were below the cut-off level of “appropriateness” set at 30% on a consensus basis by North American guidelines.4 5 We had an excess number of hospitalisations for respiratory diseases, heart failure, some subsets within the gastrointestinal/hepatobiliary group and for infections or metabolic causes. Although these conditions are considered generally suitable for short observation,4–7 9 our experience showed that the 24 h time limit was unrealistic for some of them including exacerbations of chronic obstructive pulmonary disease or pneumonia (only young patients with attacks of asthma could be discharged within 24 h). Also, the observation and conservative treatment of upper gastrointestinal bleeding deemed on clinical grounds alone to be at low risk of complications was unsuccessful since many patients had to be admitted after endoscopy. There are no published data on the management of intestinal obstruction (by postoperative adhesions) in the observation wards. In fact, our findings discourage this practice since we found that even patients with the mildest type of obstruction could not be discharged in more than 40% of cases. The inability to apply the guidelines and exclusion criteria correctly during the initial evaluation was the main reason for abnormal hospitalisation rates in the remaining cases. This problem was particularly evident for biliary pain (compounded sometimes by the imprecision of ultrasonography to identify the complications) and for heart failure (improper selection of patients with acute severe left ventricular failure rather than with mild congestive symptoms needing only extra diuretic therapy). These results undoubtedly will lead us to reconsider critically the opportunity and modalities of observation for these categories of patients.

The use and availability of some diagnostic procedures deserve some comments. The exercise tolerance tests requested to the cardiology service were the most important bottleneck although, in 2007, more tests were performed before discharge and many more patients with pain of possibly cardiac ischaemic origin were formally evaluated in the SOU without missing any acute coronary syndrome. Doppler ultrasound was not feasible on weekends and holidays, so that some patients theoretically eligible were not admitted to the SOU for this reason. On the other hand, ultrasonography carried out by some of the emergency physicians was found to be a double-edged sword since it allowed a timely but sometimes unreliable evaluation of the biliary tree.

The risk of reattendance (and of hospital admission) after discharge was significantly lower in 2007, confirming the long-lasting benefits of the intervention except for patients with recurrent atrial fibrillation. In fact, our attempt to modify the attitude of clinicians towards a policy of “rate control” instead of “rhythm control” was largely unsuccessful.

CONCLUSION

We do not offer absolute proof that the multipurpose SOU was effective as there were practical reasons which prevented us from carrying out a traditional randomised controlled trial to test such a hypothesis. However, our study showed that adequate staffing, clear admission criteria, goals, time limits, continuous clinical supervision and locally adapted clinical pathways help to streamline the flow of patients, with better results than in an SOU without these features. Since we measured the combined effect of clinical policies and senior supervision on the performance of the SOU, we are unable to discriminate their separate effects, even though we had the impression that neither of these two alone would have been sufficient to produce the results observed. With regard to external validity, this study reflects to a large degree the syndromes or diseases treated so far in similar units or eligible for observation on a consensus basis, with some exceptions. Alcohol-related and social problems, for example, were not represented in this study, reflecting a compassionate rather than accountable and standardised approach, so we cannot anticipate the role of SOU for these patients. In any case, the selection of cases suitable for short observation needs a critical appraisal of the time limits allowed, which may be too tight for some subgroups of patients. In fact, the differences rather than the similarities between the SOU and dedicated wards where a more prolonged stay is allowed (eg, MAU, MAPU) must be taken into account. While the former represents an attempt to avoid unnecessary hospital admissions and unsafe discharges, the latter are directed mainly to optimise the length of stay and clinical pathways of so-called “fast track” hospitalisations. Our results cannot therefore be extrapolated to these forms of health care, particularly for the selection of suitable cases which may differ substantially. With this caveat, we believe that our experience can be successfully replicated in many hospitals in Italy and elsewhere.