Irish Journal of Medical Science

, Volume 179, Issue 2, pp 255–258

Frequency and risk factors associated with emergency medical readmissions in Galway University Hospitals


  • J. Gorman
    • Regional Health OfficeMerlin Park University Hospital
  • A. Vellinga
    • Department of General PracticeNational University of Ireland
  • J. J. Gilmartin
    • Department of Respiratory MedicineMerlin Park University Hospital
    • Department of Geriatric Medicine, Unit 4Merlin Park University Hospital
Original Article

DOI: 10.1007/s11845-009-0452-z

Cite this article as:
Gorman, J., Vellinga, A., Gilmartin, J.J. et al. Ir J Med Sci (2010) 179: 255. doi:10.1007/s11845-009-0452-z



Unplanned readmissions of medical hospital patients have been increasing in recent years. We examined the frequency and associates of emergency medical readmissions to Galway University Hospitals (GUH).


Readmissions during the calendar year 2006 were examined using hospital in-patient enquiry data. Associations with clinical and demographic factors were determined using univariate and multivariate analyses.


The medical emergency readmission rate to GUH, after correction for death during the index admission, was 19.5%. Age 65 years or more, male gender, length of stay more than 7 days and primary diagnoses of chronic obstructive pulmonary disease, myocardial infarction, alcohol-related disease and heart failure during the index admission were significantly associated with readmission in univariate and multivariate analyses.


The medical emergency readmission rate in GUH is comparable to other acute hospitals in Ireland and Britain. Further evaluation is needed to estimate the proportion of readmissions that are potentially avoidable.


Hospital readmissionInternational classification of diseasesChronic diseases


As in other Western countries, emergency medical admissions have been increasing in Ireland during recent years and are major contributors to the frequent overcrowding in emergency departments in acute hospitals [13]. Unplanned readmissions after discharge from hospital have also been increasing [4]. Reducing readmissions, especially those that may result from poor quality care or from overhasty discharge, is an important health policy goal [5]. Epidemiological studies of medical readmissions can give a better understanding of the factors that are associated with emergency readmissions. This may facilitate targeted interventions to reduce unnecessary readmissions, reducing the pressure on acute hospitals, whilst also saving money and improving the quality of care.

The hospital in-patient enquiry (HIPE) system, which operates in all acute public hospitals in Ireland, provides a national database of coded discharge summaries [6]. Using HIPE data, we examined the profile of emergency medical readmissions in Galway University Hospitals (GUH), the largest teaching hospital in the west of Ireland.


Study setting

We examined data relating to emergency medical patients admitted to Galway University Hospitals (GUH) in 2006. GUH is the collective name for University College Hospital Galway (UCHG) and Merlin Park University Hospital (MPUH), located about three miles apart. UCHG is the main receiving hospital for all emergency admissions via a dedicated emergency department. Emergency medical ‘on take’ days alternate between the two hospitals. Ethical approval was obtained from the Research Ethics Committee of GUH.


The HIPE coding system uses the International classification of diseases (10th edition), which groups diseases or symptoms/signs into 22 codes. The computer-based HIPE system collects clinical, demographic and administrative data from the Patient Administration System (PAS) in each hospital [6]. Principal diagnoses (and up to 19 secondary diagnoses) are collected. Since 2005, the coding allows the specific selection of medical emergency admissions only. The specialities of haematology, neurology and oncology were excluded from the study because these specialities do not partake in the ‘on take’ medical emergency admission system.

Patients in UCHG and in MPUH have different hospital numbers. Four variables (surname, date of birth, gender and admission date) and visual inspection of the database were used to accurately identify readmissions and transfers across the two sites.

The focus in this study was on the primary diagnoses. Several common diseases, e.g. respiratory tract infections, have multiple codes and these were grouped together to allow further analysis.

Statistical analysis

The data were analysed in the first instance using descriptive techniques. The primary outcome measure in this study was the occurrence of any readmission during the year 2006. Readmission within 30 days was examined as a secondary outcome. The associations between demographic factors and the most common primary diagnoses (those occurring in 100 patients or more) and readmission were examined with univariate analyses. Age and length of stay were analysed using non-parametric methods due to their skewed distribution.

Logistic regression was used to determine the multivariate relationship between readmission and those factors that were significantly related to readmission in univariate analyses. A conservative significance level of P < 0.1 was used to select factors for consideration in the multivariate model. Also, because the existing literature suggests that age and sex are important determinants of readmission, it was decided that these variables should be forced into the model irrespective of the univariate results. A backward selection model was used otherwise to determine the final model. The Nagelkerke R2, which ranges between 0 (no association) and 1 (perfect association), was used to provide a guide to the power (effect size) of the model [7]; this can be regarded as a crude approximation of the ordinary least squares R2 in linear regression, which measures the proportion of variance in the outcome explained by the regression model. Data analysis was performed using SPSS version 15.


In the calendar year 2006, 5,280 patients (51.2% male) had 6,854 medical emergency admissions to GUH (2,944 admissions to MPUH and 3,910 to UCHG). The median age was 65 years [interquartile range (IQR) 46–77]; 49.4% were less than 65 years, 28.2% were 65–80 years and 22.2% were more than 80 years. The median length of stay (LOS) was 5 days (IQR 3–10); 64.4% stayed for 7 days or less, 31.5% stayed for 8–30 days, 3.6% stayed for 31–90 days and 0.5% for more than 90 days.

Overall, 987 (18.7%) patients had more than one admission in 2006: 686 (13.0%) were admitted twice, 191 (3.6%) three times, 63 (1.2%) four times and 47 (0.9%) five times or more (to a maximum of 13 admissions). Excluding the 219 patients who died during the index admission, the readmission rate during 2006 was 19.5%.

Table 1 shows the patient and disease factors associated with readmission. Patients who were readmitted were significantly older (median age 72 years, IQR 57–81 years) than those not readmitted (63 years, IQR 42–77 years) (Mann–Whitney U, P < 0.0001). The LOS during the index admission was significantly longer for patients who were readmitted [median 7 (IQR 4–13) vs. median 5 (IQR 2–9), Mann–Whitney U, P < 0.0001]. Male gender was also associated with increased likelihood of readmission. Of the 12 most common primary diagnoses, heart failure, chronic obstructive pulmonary disease (COPD), myocardial infarction and alcohol-related disease were associated with significantly higher readmission rates (Table 1), whilst respiratory disease, syncope, ischaemic heart disease, stroke or transient ischaemic attack, epilepsy, atrial fibrillation, urinary tract infection and chest pain were not (data not shown).
Table 1

Patient and disease factors associated with readmission



Readmitted (%)











 <65 years




 65–80 years



 >80 years




 <8 days




 8–30 days



 >30 days











Myocardial infarction








Alcohol related








Heart failure








COPD chronic obstructive pulmonary disease, LOS length of stay

The logistic regression model for readmission during 2006 shows that older age, male gender, prolonged initial LOS and diagnoses of heart failure, COPD, myocardial infarction and alcohol-related disease were independent predictors of readmission (Table 2). However, the overall predictive power of the model was relatively poor with a Nagelkerke R2 of only 0.06.
Table 2

Logistic regression models predicting readmission


Calendar year model



1.30 (1.08–1.56)

1.20 (1.0–1.40)

LOS >7 days

1.55 (1.34–1.80)

1.43 (1.14–1.81)

Age ≥65 years

1.74 (1.49–2.0)

1.31 (1.03–1.66)


2.99 (2.17–4.12)

3.16 (2.11–4.74)

Heart failure

1.70 (1.12–2.58)

Myocardial infarction

1.63 (1.10–2.42)

Alcohol related

2.96 (1.91–4.57)

Data are odds ratio (95% CI)

Calendar year readmission model chi square = 204.3, df 7, P < 0.0001; Nagelkerke R2 = 0.06

30 day readmission model chi square = 56.4, df 4, P < 0.0001; Nagelkerke R2 = 0.03

The readmission rate within 30 days of discharge was 6.7% (354 patients). Older age, male gender, longer LOS and diagnosis of COPD were independent predictors of readmission within 30 days in the multivariate regression model (Table 2). Again, the predictive power of this model was poor with a Nagelkerke R2 of 0.03.


The readmission rate in Acute Medicine to Galway University Hospital, for the year 2006, was 19.5%. Older age, male gender, longer initial length of stay and diagnoses of heart failure, COPD, myocardial infarction and alcohol-related disease during the index admission predicted readmission in both univariate and multivariate analyses.

Despite differences in how the studies were conducted, our results were comparable to those reported from a study of readmissions in St. James’ Hospital, Dublin [8]. In their study of 4,050 patients admitted in the calendar year 2002, the readmission rate was 15.0%. One possible explanation for the slightly higher readmission rate in Galway University Hospital compared with St James Hospital is that the former is the only major hospital in the catchment area and readmissions may be less likely to ‘leak’ into other hospitals than in South Dublin, which has three major hospitals. Our results are also comparable to those reported from a prospective study in a district general hospital in Manchester where the 28-day readmission rate was 7.2% and the 12-month (not calendar year) rate was 23% [9].

Male gender, a longer initial admission and diagnosis of heart failure, COPD and alcohol abuse have been shown to predict readmission in this as in many other studies [5, 1012]. We, like others [10], found that older age predicted readmission, but this has not been a consistent finding in all studies [12, 13]. It seems likely that advanced age is often a surrogate marker for multiple comorbidities, as well as for socioeconomic deprivation, and isolation and that correction for the latter factors reduces the importance of age per se as a predictive factor [5].

As in other studies [7, 1416], this study used data from the HIPE database linked to the hospital record. There have been concerns over the limitations and accuracy of such data and this approach does not allow examination of important socioeconomic and demographic factors. Our focus on the primary diagnosis meant that we did not take potentially important comorbidities into account. However, there is evidence that the accuracy of coding is significantly more accurate for the primary than for secondary diagnoses [17]. Although the multivariate logistic model generated from the data identified significant predictors of readmission and the model satisfied the goodness of fit criteria, the overall predictive power of the model was relatively poor.

Our data did not allow us to examine the outcomes with readmission in a meaningful way. However, other reports suggest that early unplanned readmission is related to poor outcomes including increased mortality during the 2 years after the index admission [18]. Poorly documented discharge planning, increased temperature, intravenous fluids on the day of discharge or unaddressed abnormal test results at discharge were related to an increased subsequent mortality in another study [19]. Furthermore, interventions aimed at reducing readmission rates have been associated with significant reduction in mortality [20].

One tempting approach is to consider that early unplanned readmission reflects a failure of the care given on the index admission and, hence, that readmission rates may represent a useful quality indicator. There is indeed some evidence to support this. For example, a meta-analysis of 16 studies found that the risk of early readmission (within 31 days) was increased by 55% when care was of relatively low quality, that is, substandard or normative instead of normative or exceptional [21]. However, other studies suggest that the reality is much more complex [5, 22]. For example, DesHarnais et al. [23] ranked 300 hospitals on three risk-adjusted indices of hospital quality: mortality, readmissions and complications. They found no relationship between a hospital’s ranking on any one of these indices and its ranking on the other two indices.

The balance of evidence suggests that readmission rates, uncorrected for confounding medical, social and hospital factors, are a poor guide to quality of care [5, 20, 24].

Whilst between 9 and 48% of medical readmissions can be judged as due to health-care factors, such as suboptimal health or social care within the hospital or the community, other significant contributors include unavoidable social factors, patient factors (such as poor compliance) and disease factors (such as natural progression or recurrence of disease) [5]. For many patients with chronic diseases, in particular, readmissions may be unavoidable and, in fact, represent optimal care.

Despite these reservations and the need for caution in interpreting our and other reports on readmissions, examination of readmission rates may give some indication of which patient groups or medical conditions may require extra resources or innovative approaches to minimise readmissions. In particular, high readmission rates of patients with well-defined chronic conditions may lead to identification of remediable process or quality of care problems. Further research within the Irish health-care system is necessary to identify the proportion of readmissions that are due to health-care factors and are potentially preventable. Similarly, the identification of heart failure, COPD and alcohol abuse as important predictors of readmission supports the importance of recent developments in improving the management of these common chronic conditions, including, for example, development of heart failure clinics and pulmonary rehabilitation programmes.

Conflict of interest statement


Copyright information

© Royal Academy of Medicine in Ireland 2009