INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

It is common to lament the fact that we spend too much money providing intensive care for people in their last months of life and not enough on curative or preventive care (1). Such concerns focus discussions of resource allocation on the intensive care unit (ICU), where such poor allocation decisions are thought to take place (2). Discussions of how to remedy the situation vary. Some focus on patient's preferences, hypothesizing that people don't really want such care but that it is forced upon them by physicians who are driven by technological imperatives (3). The solution to this problem would be to empower patients to refuse therapy by encouraging the use of advance directives (4).

Others suggest that resource allocation could be improved through clinical epidemiological data about the prognosis for individual patients or groups of patients (5). Such data may rely on physiologic variables such as illness severity scores or on status variables such as age (6) or birth weight (7). In either case, decisions to forgo potentially life-sustaining treatment that are not based on patient's preferences require comparisons of the relative cost and efficacy of different treatments for different patients (8).

Babies of extremely low birth weight and elderly adults require expensive and scarce resources, and both have a relatively poor prognosis for survival if they require intensive care. Thus, proposals for rationing often target one or both of these groups (9).

While both newborns and the elderly clearly require costly medical care, we thought that differences in the patterns of hospital care for the two populations might be relevant to resource allocation decisions. In particular, we suspected that neonates who die generally do so early in their hospital course, while those who survive for a few days generally have an excellent prognosis. By contrast, we suspected that elderly patients who require intensive care are more likely either to recover quickly or go on to a slow and lingering death in the ICU. If our suspicions were true, overall mortality rates would not accurately reflect the cost-effectiveness of treatment for the two populations. Although overall mortality rates might be higher in the NICU, the care might nevertheless be more cost-effective. To test this intuition, we examined mortality patterns in our NICU and compared them with patterns in our adult general medicine ICU (GMICU) and a cardiac ICU (CICU).

We chose medical ICUs, rather than surgical ICUs, for two reasons. First, we thought that MICUs were more comparable to the NICU, where patients are generally admitted for medical rather than surgical problems. Second, we recognized the continuum between the surgical recovery room and the surgical ICU and thought that institution-specific policies about the utilization of these two sites for intensive care of postoperative patients might make entry criteria less generalizable. Our goal was to see how intensive care of patients at the extremes of life compared both in overall survival as well as in the resources allocated to survivors and nosurvivors.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

For the calendar year 1993, we reviewed the medical records of all patients admitted to the NICU (n = 827), adult GMICU (n = 570), and adult CICU (n = 750) at The University of Chicago Hospitals. For each patient, the following items were determined: birth weight (for NICU patients), age (for adult patients), date of ICU admission, date of death (for nonsurvivors) or discharge from the ICU (for survivors), and length of ICU admission. We did not examine survival until hospital discharge.

Infants admitted to the NICU are admitted because of prematurity and its complications. Thus, we did not record admitting diagnosis in the NICU. For adult patients, we determined the primary diagnosis that led to ICU admission. We analyzed outcomes for all patients and for the subset of patients who received mechanical ventilation.

Four extremely premature infants were considered "not viable" at birth and were admitted to the NICU but were neither intubated nor resuscitated. These infants were included in the analyses of the overall NICU population but were excluded from the analysis of the patients who received mechanical ventilation.

Statistical comparisons between neonatal patients (NICU) and adult ICU patients (combined GMICU and CICU) were performed using either Student's unpaired t test for continuous variables or the chi-square test for categorical variables.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Table 1 shows the diagnoses for patients admitted to the ICU at the University of Chicago. Of the babies admitted to the NICU, most were low-birth-weight babies with problems of prematurity or babies with congenital anomalies.

Figure 1A displays mortality as a function of birth weight for 827 infants admitted to our NICU. Figure 1B displays similar data for 1320 adults admitted either to our GMICU or CICU, plotting mortality against age rather than birth weight. The figures show mortality for all patients and for patients who required mechanical ventilation. In the NICU, decreased birth weight was associated with increased mortality. For the adults, increased age was associated with increased mortality. Although overall mortality in the NICU (66/827; 7.9%) was significantly lower than in the adult ICUs (219/1320; 16.5%) (p < 0.001), mortality for the smallest newborns (< 750 g; 51% mortality) was higher than for the oldest adults (> 84 years; 30% mortality) (p = 0.05). The mortality curve for ventilated NICU patients closely approximates the curve for NICU patients overall. In contrast, mortality for ventilated adults is much higher than for adults overall, suggesting that ventilation is a more discriminating measure of illness severity in adults than neonates. Ventilated adults of any age were significantly more likely to die than ventilated newborns of any weight (p < 0.05 for all comparisons).

View larger version (30K): [in this window] [in a new window]   View larger version (48K): [in this window] [in a new window]   Figure 1.   (A) Mortality versus birth weight for NICU patients. (B) Mortality versus age for ICU patients.

Table 2 compares the length of stay for newborns who died with the length of stay for adults who died. From this table, it is clear that newborns who die tend to die early, while adults who die tend to die after a more prolonged ICU stay. The disparate tempo of dying in the NICU and ICU was more pronounced in the highest risk groups. For newborns with birth weights less than 751 g, 60% of those who died did so within the first 2 d of life. Conversely, for adults, only 38% of those who died did so within the first 2 d. For those over 84 y of age, this number was even lower (31%).

This phenomenon is elaborated in Figure 2, where survival to hospital discharge is depicted as a function of day of ICU stay for both the neonatal and adult ICU populations at greatest risk to die. Considering the smallest neonates first, overall survival was approximately 49% for all infants with birth weights < 751 g who were alive on Day of Life (DOL) 1 (that is, all infants included in the study population). During the first NICU day, seven of these patients died, and none of the remaining patients were transferred out of intensive care. Consequently, for the 32 infants remaining in the ICU on DOL 2 the likelihood of ultimate survival was enhanced (increasing from 49% to 59%). This phenomenon (early death > early discharge, leaving a healthier and healthier resident NICU population) continued for the first several days of life, leading to a progressively greater likelihood of survival (approaching 80% by ICU day 10) with each passing day in the NICU.

View larger version (50K): [in this window] [in a new window]   Figure 2.   Survival versus day of ICU stay for patients at greatest risk to die.

Precisely the opposite tendency was noted for adults. Overall, survival for adults 85 yr and older was 70% on ICU day 1. During the first ICU day, 4 adults died (22% of all deaths in this age group), while 17 were transferred out of the ICU (41% of all live discharges). Consequently, for the 39 patients in this population who remained in the adult ICU on day 2, the likelihood of survival was diminished compared with the day 1 population (falling from 70% to 64%). This phenomenon (early discharge > early death, leaving a sicker and sicker resident ICU population) continued for the first 2 wk of ICU hospitalization. Consequently, with each passing day in the adult ICU, the remaining population was increasingly likely to die (survival falling below 30% by ICU day 10).

These divergent survival patterns have a profound impact on resource utilization in the two ICUs. Because neonates who die tend to die quickly, most of the bed-days in the NICU are devoted to patients who ultimately survive. Because adults who die tend to have more prolonged courses in the ICU before succumbing, a much higher percentage of ICU bed-days are allocated to patients who will ultimately die. These proportions are shown in Figure 3 for neonates, adults, and adults who required mechanical ventilation, stratified by mortality risk (birth weight or age). Overall, the percentage of ICU bed-days devoted to nonsurviving adults (28.8%) was significantly larger than the percentage of NICU bed-days devoted to nonsurviving babies (7.8%). Even among babies at greatest risk to die (birth weight < 751 g), the percentage of NICU bed-days allocated to nonsurviving infants was less than 20%. In contrast, for the oldest ICU patients (> 84 yr old) this value exceeded 50%; for ICU patients > 84 yr old who required mechanical ventilation, the percentage of ICU bed-days allocated to nonsurvivors approached 90%.

View larger version (33K): [in this window] [in a new window]   Figure 3.   Bed-days allocation to nonsurvivors in the NICU and ICU.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Intensive care is expensive. Patients who require such care have been the focus of critical ethical analyses that seek fair ways to allocate scarce resources. In such analyses, the principle of triage is often invoked in order to target resources to the patients with the best chance of survival and to avoid the allocation of resources for patients with unacceptably low survival rates.

Comparisons among ICU populations have often used overall survival statistics. Our data suggest that these ICU survival statistics miss the mark, at least in the context of resource allocation and distributive justice. Instead, discussion might better be framed in terms of number of ICU bed-days (a proxy for dollars) utilized by survivors (or nonsurvivors) divided by total number of ICU patients bed-days/dollars expended. As an example, using the mortality criterion, neonatal intensive care for babies with a birth weight < 750 g (mortality 51%) would appear less effective than intensive care for adults over 84 yr of age (mortality 30%). However, for the same populations, only 19% of NICU bed-days were expended on nonsurviving infants, contrasted with 54% of ICU bed-days allocated to nonsurviving adults. Even for populations at lower risk of dying, the fraction of ICU resources expended on nonsurviving adults consistently exceeded the percentage of NICU resources spent on nonsurviving newborns.

In an attempt to minimize the possibility that variations in admitting criteria to ICUs might confound our results (i.e., independent of how many "good" 1,800-g premature infants or "healthy" adults with possible pulmonary embolus were admitted to the ICUs as opposed to a "floor" or "intermediate" bed), we identified the subset of each ICU population requiring mechanical ventilation (a rough measure of illness severity) and determined mortality and length of stay of these subpopulations as well. Our choice of respiratory failure/mechanical ventilation as a comparison criterion across adult and neonatal ICU populations was spurred by the recognition that at our institution (as, we suspect, at many institutions) the initiation of mechanical ventilation mandates ICU care. When ICU and NICU populations were adjusted for this illness severity criterion (that is, when the study groups were limited to patients with respiratory failure severe enough to require assisted mechanical ventilation), the distinction between adult and neonatal ICU resource utilization for nonsurvivors was accentuated in all risk categories. Adult ICU patients with respiratory failure who died consumed many times more ICU resources before their death than did NICU patients.

The poor survival of tiny premature infants has at times been used to argue that such infants should not be treated at all (10). Our data suggest that although survival for infants of extremely low birth weight may be very unlikely, the cost to society for the care of nonsurvivors is relatively low. Concerns about distributive justice in allocating resources to broad classes of patients should demand that resources be targeted to those patients who will survive. Over 80% of dollars spent on NICU care for newborns even as small as 750 g or less will be spent on infants who ultimately survive, compared with less than 50% of dollars spent on ICU care for patients 85 yr or older and less than 20% of dollars spent on ICU care for patients 85 yr or older who require mechanical ventilation.

There may be a number of reasons for these differences. The first may be physiologic. Extremely premature babies often develop rapidly progressive pulmonary failure and intracerebral hemorrhages early in life. Those babies who make it through the first days of life without these complications are much less likely to develop them later on. These physiologic facts lead to the conventional wisdom that premature babies "declare themselves" and to a clinical policy of initiating treatment on most babies in order to see which will do well and which will develop life threatening complications.

A second possible explanation, not examined by our study, is that neonatologists make decisions to withhold treatment more frequently or more judiciously than do physicians in adult ICUs. This seems unlikely. Available data suggest that decisions to withhold or withdraw life-sustaining treatment are more common for elderly patients in adult ICUs than for extremely low-birth-weight babies in neonatal units (11, 12).

Finally, our data might be interpreted as showing that neonatal intensive care is overused for many babies. If survival rates for neonates are high after the first few days of life, perhaps they do not need the same intensity of services. This is probably partly true and has been recognized by some managed care and insurance plans which reimburse NICUs more for care in the first days of life than for care later. Nevertheless, even after the first days of life, most premature babies continue to need many of the services that can only be provided in NICUs, and their excellent survival rates would likely be lower without such care. This issue seems to call for fine tuning of actuarial calculations in order to "cost out" the care of older premature babies who are medically stable and growing, rather than a change in practice patterns leading to discharge of these infants from appropriate clinical settings.

Although there may be many legitimate concerns about justice and ethics in the NICU, our data suggest that undue expenditure of society's resources prolonging the dying of extremely small infants is not among them. By contrast, intensive care for the elderly involves a far greater proportional expenditure of money toward those who will not survive. To the extent that concerns about distributive justice drive allocation decisions in ICU care, it would seem more justifiable to ration intensive care for the very old, not for the very young.