Patients’ Characteristics Associated With Readmission to a Surgical Intensive Care Unit

In this prospective observational cohort study, all 1979 consecutive surgical patients admitted to the ICU at St Elisabeth Hospital, Tilburg, the Netherlands, a 673-bed teaching hospital, were included in the cohort. The duration of the study was 61 months, from January 1995 until February 2000. The 32-bed ICU admits medical and surgical critically ill patients and is staffed by critical care specialists with an internal medicine, anesthesiological, or surgical background. The study was part of a continuous quality improvement program (1995–2000) and was approved by the local medical ethics committee.

Because the surgical population is a distinct ICU population, in both admission profile and outcome, patients in specific surgical subspecialties—trauma, vascular, gastrointestinal, oncological, and general—were the focus of the study. Exclusion criteria were age less than 18 years, readmission more than 30 days after first ICU discharge, and gynecological and nontrauma neurosurgery. Demographic data were collected on all patients: sex, age, type of surgical classification, severity of illness as indicated by the score on the Acute Physiology and Chronic Health Evaluation (APACHE) II¹⁶  and the Simplified Acute Physiology Score (SAPS) 3¹⁷ ,¹⁸  any preadmission disease (as scored in the chronic health points of the original APACHE II and comorbid conditions), elective or emergency admission, whether the admission was postoperative, type (indication) of first ICU admission, and lengths of first ICU and hospital stay. The APACHE II score was calculated in accordance with the original definitions (specific coefficients for diseases) by using the worst physiological values as measured during the first ICU day.¹⁶ 

Indications for admission were categorized as low-risk monitoring (admission for monitoring and observation), high-risk monitoring (patients in stable condition but with high risk of complications and need for extended care), and active treatment (patients with unstable conditions and need for continuous medical intervention). Initial ICU length of stay was defined as the period from the day of the first admission until the day of discharge from the unit. Length of (initial) hospital stay was defined as the period from the day of hospital admission until definitive discharge from the hospital, including the initial ICU admission and eventual readmission periods.

Readmission to the ICU was defined as a return within 30 days (early or late readmission, ≤48 or >48 hours, respectively), internationally defined as ICU readmission after discharge from original ICU admission. The mean hospital length of stay was 25 days.

Survival was determined by reviewing the hospital’s electronic patient medical record. If a patient’s death could not be confirmed by the database, the patient’s general practitioner was consulted. If a date of death could not be found, the patient or his or her relatives were consulted. Follow-up of each (surviving) patient was continued until February 2006. All patients were followed up for at least 6 years (mean, 8.4 years; SD, 1.4 years). A total of 52 patients were lost during follow-up (the follow-up rate was 97%). Quality of life was measured in all patients (those readmitted and those with a single admission) who were alive at the end of follow-up, February 2006. The EuroQol-6D questionnaire¹⁹ –²⁴  (EQ-6D) was used to measure quality of life. The first EQ-6D outcome (EQ-utility score) was obtained by using the United Kingdom EQ-5D index tariff for all possible health states defined by the EQ self-classifier.²⁵ –²⁷  The cognitive dimension was ignored (not yet available for analysis). This index tariff links a single index value to each hypothetical health state.²⁵  The second outcome was the percentage of problems for each dimension. The outcome categories (no problems, mild problems, and severe problems) of these dimensions were dichotomized to no problems and problems (mild or severe).

A total of 16 patients had a missing value on either the APACHE II or SAPS 3. For these patients, the mean value was imputed by using linear regression analysis based on age, sex, and type of surgery. This method differs from the approach of Knaus et al,¹⁶  who used normal values for missing data.

The incidence of ICU readmission within 30 days after first ICU discharge was measured. Data on patients who died during first ICU stay or 30 days or less after the original ICU discharge were excluded because these patients might have died before readmission. Subsequently, multivariable logistic regression analysis was used to determine the independent association of patients’ characteristics with ICU readmission. Age, APACHE II score and SAPS 3 value were included as linear parameters, after cubic spline analysis confirmed the linearity.²⁸ 

Short-term, ICU, and long-term survival were compared between patients who were readmitted and those who were not by using the Pearson χ² test. The quality-of-life outcome was compared by using the nonparametric Mann-Whitney test (EQ-utility score variable) and the Pearson χ² test (percentage of reported problems, dichotomized value). Cox regression modeling was used to generate survival curves.

Values of P < .05 were considered significant. Statistical analysis was performed by using SPSS 15.0 (IBM SPSS) for Microsoft Windows XP Professional, version 2002, service pack 2.0.

The baseline characteristics of the total sample are given in Table 1. Of these 1979 patients, 297 died during their initial ICU admission or within 30 days after discharge from the ICU. Of the 1682 surviving patients, 141 (8%; 95% CI, 5–14) were readmitted to the ICU within 30 days after discharge. Readmission rates were 14% (95% CI, 11–19) for gastrointestinal(35/244), 10% (95% CI, 10–13) for vascular (55/538), 9% (95% CI, 8–11) for oncological (29/316), 4% (95% CI, 3–5) for trauma (15/422), and 4% (95% CI, 4–7) for general surgery (7/162) patients. Most of these patients (83%) were readmitted within the first 2 weeks (Figure 1). A total of 28 patients were readmitted within 48 hours (1.7% of the total cohort of 1682 patients; 20% of all patients readmitted). Causes of readmission were respiratory failure (68 patients, 48%) cardiac problems (16%), sepsis (14%), reoperation (11%), and other reasons (10%).

Compared with patients who were not readmitted, patients who were readmitted were older, mostly had vascular (39%) or gastrointestinal surgery (25%), had a significant higher initial severity-of-illness scores (APACHE II: mean 12; SD, 6; vs mean, 11; SD, 6; P = .007; SAPS 3: mean, 46; SD, 12 vs mean, 42; SD, 13; P = .003), and had more comorbid conditions (14% vs 8%; P = .005). In addition, more patients who were readmitted required more intensive care at the initial admission (low-risk monitoring: 49% of readmitted vs 63% of not readmitted, P = .001; high-risk monitoring: 34% vs 27%, P = .09; active treatment: 17% vs 10%; P = .004; Table 1). Patients readmitted within 48 hours were significantly older, had more comorbid conditions, had higher initial severity-of-illness scores (SAPS 3), and more often received mechanical ventilation than did patients readmitted more than 48 hours after the latter patients’ initial ICU admission. In the various surgical subspecialties, compared with patients who were not readmitted, only readmitted patients in the vascular (20% readmitted vs 10% not readmitted; P = .005) and oncological (26% vs 6%; P = .001) surgery groups had significantly more preadmission comorbid conditions. As indicated in Table 2, most surgical subspecialties (gastrointestinal, vascular, and oncological), higher level of care (high-risk monitoring and active treatment needed during initial admission), and need for mechanical ventilation were significantly independently associated with readmission.

Short-term (ICU) mortality did not differ significantly between patients who were readmitted (10%) and those who were not (11%). Irrespective of its cause, the long-term mortality rate (6 years after ICU discharge) was significantly higher for patients who were readmitted (70% vs 42%; P< .001; Figure 2). Long-term mortality according to different surgical classifications was significantly higher for the gastrointestinal and vascular patients who were readmitted.

Of the 141 patients who were readmitted, only 43 survived until the end of the follow-up period. Quality of life (EQ-6D) was measured for 32 of these 43 patients (74%). Quality of life was also measured in 575 of the 886 patients (65%) who were not readmitted. As shown in Table 3, the median EQ-utility score and the percentage of reported problems did not differ significantly between the 2 groups of patients.

During a 5-year study period, 141 of 1979 surgical ICU patients (8%) were readmitted to the unit within 30 days after discharge.

Compared with patients who were not readmitted, those who were readmitted were older, had a higher prevalence of preadmission disease (comorbid conditions), had worse initial SAPS 3 values, and required more intensive treatment than monitoring only during the initial admission. Our results did not indicate any predictive variable for readmission at the beginning of the initial ICU admission. We found an independent association with readmission for all surgical classifications (except general surgery), admission type, and the need for mechanical ventilation. Most readmissions were due to respiratory distress and occurred within the first 2 weeks after discharge from the ICU. A total of 28 patients (1.7%) were readmitted within 48 hours of discharge. These patients were significantly older, had more comorbid conditions, had higher initial severity-of-illness scores, and received mechanical ventilation more often than did patients readmitted more than 48 hours after the latter patients’ initial ICU admission. More than 80% of readmitted patients had been discharged from the ICU for a period longer than 48 hours. Except for trauma and general surgical patients, patients who were readmitted had significantly higher severity-of-illness scores at readmission than they did at their initial admission.

Patients who required readmission had a significantly worse long-term outcome than did patients who were not readmitted. Morbidity of the surviving readmitted patients was less than an age- and sex-matched general population norm.¹⁹ ,²⁰  The EQ-utility score did not differ significantly between patients who were readmitted and those who were not.¹⁹ 

Our ICU readmission rate was comparable to the 3% to 13% (surgical ICU population) and 5% to 10% (mixed ICU) rates reported by other investigators.² –⁴ ,⁶ –¹⁶  The causes of readmission in our study are in accordance with the causes in many earlier reports.² –⁵ ,⁶ –¹¹ ,¹⁶ ,²⁹  In general, respiratory distress is the main cause for readmission to intensive care. Unlike Nishi et al,⁹  we did not detect a decrease in readmissions for respiratory failure during the study period (Figure 2). Although the readmission rate because of sepsis or reoperation varied from year to year during the study, numbers were low for all groups.

Readmission to the ICU is an unexpected, unfavorable event for patients, patients’ family members, and health care workers and is associated with a worse outcome than the outcome for patients who are not readmitted. In our study, patients were sicker at readmission than at their initial admission, as indicated by the SAPS 3 value. Our hospital mortality rate (16%) is lower than the 20% to 46% rates reported by other researchers.⁴ –⁶ ,⁹ –¹¹ ,²⁸ ,²⁹  Nevertheless, readmitted patients do have a longer hospital admission and a significantly worse long-term survival than patients who are not readmitted. Other research⁴ ,⁷ –⁹ ,¹¹ –¹⁵ ,²⁹ ,³⁰  has shown that early ICU readmission is associated with morbidity and mortality rates that are higher than the rates of patients who are not readmitted.

The question arises then of whether or not readmission could be an important quality indicator. Readmission might indicate that the readmitted patients were discharged prematurely. If the patients were discharged too soon, the incidence of early readmission must be high. Of all the patients readmitted in our study, 20% were admitted early (≤48 hours after ICU discharge). This percentage is slightly lower than published rates of 22% to 30%.³ ,⁵ ,⁷ ,²⁹  Moreover, a performance quality indicator requires a certain number of patients.² ,⁶  Our early readmission rate of nearly 2% (of the total study population) is probably not sufficient for a performance quality indicator. The relation between quality of care and late readmission (most of our study population) is questionable. Late readmission could be an indication of pathophysiological disturbance in a patient with an underlying disease or comorbid condition. Although early readmission may be related to the initial cause for ICU admission, late readmissions most likely are related to factors beyond initial ICU control. Hypothetically, these readmissions are more likely to be a second ICU admission for a separate or additional problem rather than a true readmission. Nevertheless, Chen et al³  evaluated the reasons for ICU readmission in patients readmitted within and after 24 hours and found no significant difference in the cause of readmission or the quality of care between the 2 groups.

Among the various surgical classifications, patients who had gastrointestinal and vascular surgery had a worse outcome (readmission incidence and long-term mortality) than did patients who had other types of surgery. Hypothetically, trauma patients are in better health before ICU admission than are patients with gastrointestinal or vascular problems. The cause of ICU admission (gastrointestinal and vascular patients) is not the same as comorbid conditions. Moreover, trauma patients are significantly younger than are patients in the other classifications.

Our study has several limitations. First, the data are from a cohort study done in a single center. However, the ICU and hospital (short-term) mortality rates are almost identical to those in the Dutch National Intensive Care Evaluation³¹  and are comparable to the rates in international ICU studies from Finland,³²  Canada,³³  and the United States,³⁴ ,³⁵  indicating a certain generalizability of our data. Although St Elisabeth Hospital is not a tertiary referral hospital, it is a level I trauma center and acts as a regional referral center for a population of 2.1 million people. Research by Dowdy et al³⁶  indicates that the prospective cohort study is a powerful research design for assessing relationships between exposures (ie, admissions, interventions, diagnostic tests) and ICU outcomes. A second possible limitation is that some data were collected retrospectively. However, the collection of in-hospital data and the survival status of individual patients were prospectively registered, and only 52 patients (3% of our entire cohort) were lost to follow-up. Compared with data in other studies,³⁷ –³⁹  this number is low. Furthermore, the presence of preadmission disease (comorbid conditions) was determined solely by using the chronic health evaluation component of the original APACHE II. Using this process, we detected no comorbid conditions in 1787 patients in the total study population (90%); this number could be an underestimation. This finding does not accurately take into account many disease states that may predispose patients to poor outcomes. The use of a different, more detailed scoring system might have yielded more specific information on the degree of comorbid conditions for the study cohort. At the same time, we determined the severity-of-illness score only at the beginning of an admission or readmission. Several investigators⁴ ,⁵ ,¹¹ ,¹²  found that readmitted patients had a higher severity-of-illness score at the time of the patients’ original ICU discharge than did patients admitted only once. Therefore, our data would have been more valuable if we had collected the APACHE II score and the SAPS 3 value daily during the initial ICU admission. Currently, except for the study published by Lee et al,¹²  little information⁶  is available on how daily severity-of-illness scores change during ICU admission and therefore are predictive of discharge from the ICU. Lee et al concluded that the APACHE II score at the time of ICU discharge is a significant predictor of ICU readmission. Moreover, the APACHE II score has not been validated after the first 24 hours. This information may improve the ability to predict the readiness for discharge of individual surgical ICU patients and improve the efficiency of ICUs. In our study, patients who were readmitted had a worse long-term outcome than did patients who were not readmitted; however, quality of life did not differ significantly between the 2 groups. Only 141 patients were readmitted, and long-term quality-of-life information was available for only 32 patients. The lack of a significant difference might reflect a type I error due to the small sample size of readmitted patients and might lead to false-negative results, because we noticed a tendency for a worse quality of life for patients who were readmitted.

Use of early readmission as a quality indicator is difficult. Further research into this topic should include multiple centers/ICUs and comparisons of low performers with high performers, with regard to quality of intensive care treatment provided.

In conclusion, the adverse effect that readmission to the ICU has on survival appears to be long lasting. Patients who were readmitted had a worse long-term survival than did patients who were not readmitted. In addition, patients who were readmitted had worse initial severity-of-illness scores, had a higher prevalence of preadmission disease (comorbid conditions), and required more active treatment during their initial ICU admission. Possibly, the higher long-term mortality is associated with the higher prevalence of comorbid conditions and not with readmission. Although no single (ICU admission) factor was independently predictive of readmission, hypothetically, the conditions predisposing a patient to initial admission to an ICU, not the need for readmission, are responsible for the patient’s long-term outcome. Because of the design of our study, we cannot determine whether readmission is an independent prognostic factor of quality of care. The relationships between readmission to the ICU, inadequate care, premature or inappropriate discharge, and clinicians’ decision-making practices require more study.