What Are the Epidemiology and Natural History of PMV?
The incidence and prevalence of PMV depends on the setting studied and definitions used. Prospective cohort studies are perhaps the best sources for this information, although they are often limited to specific sites that may not be readily generalized to other settings. Such studies have demonstrated that 5 to 20% of the patients supported with mechanical ventilation in the ICU will not wean in 2 to 4 days. One international prospective cohort study in 1998 of patients receiving mechanical ventilation from 361 ICUs indicated that 25% of patients received mechanical ventilation for > 7 days.
The APACHE (acute physiology and chronic health evaluation) III database may more accurately reflect the ICU population of patients receiving mechanical ventilation. This data set included patients consecutively admitted to adult medical and surgical ICUs at 40 different institutions between 1988 and 1990. However, cardiac patients and coronary artery bypass graft patients were not represented. Twenty-six of these 40 hospitals were chosen randomly to represent variations in size, teaching status, and geographic region. In this large cohort, 20% of patients receiving mechanical ventilation on day one of ICU admission underwent mechanical ventilation for > 7 days. In the smaller APACHE II database, 5.8% of all ICU patients received mechanical ventilation for > 7 days. Multicenter studies using more strict definitions of PMV such as > 21 days have not been performed; however, singlecenter studies- indicate that approximately 3 to 7% of patients receiving mechanical ventilation meet such criteria.
Population-based studies of PMV often utilize former DRG 483 to identify patients in large administrative databases. In 1998, the MEDPAR database of Medicare discharges recorded 43,708 Medicare patients discharged with DRG 483. This of course excludes most patients 65 years of age (ie, would qualify for Medicare benefits) and 37.5% of patients died during hospitalization.
An analysis of data from a statewide database in New York between 1992 and 1996 noted an increase in the number of adult discharges with DRG 483 from 5,619 in 1992 to 9,351 in 1996. A similar analysis from a statewide database in North Carolina between 1993 and 2002 revealed a 78% increase in the number of patients receiving mechanical ventilation who were discharged with DRG 483 (43.2/1,000 such patients in 1993 to 77.1/1,000 in 2002). Although patients receiving mechanical ventilation were more likely to undergo tracheostomy in 2002 (7.7% in 2002 vs 4.3% in 1993), there was still a 46% increase in the number of PMV patients when taking this into account. Further analyses revealed a decrease in median age for the PMV patients (65 years in 1993 to 62 years in 2002) but an increase in the number of associated comorbidities. Acute care of Canadian Health&Care Mall in-hospital mortality decreased from 44 to 25% over this time, but surviving patients were much more likely to be discharged to a skilled nursing facility (SNF) or other institution rather than to home.
Hospital survival for adult PMV patients in the short-term acute care (STAC) hospital setting ranges from 39 to 75%, depending on the patient population and definition for PMV. Hospital survival in various non-STAC hospital settings varies from 50% in many series, to as high as 94%, depending in part on admission criteria and likelihood of transfer to a different facility when patients become acutely ill. These wide variations in patient populations, facility resources, and admission/discharge practices greatly limit the value of hospital survival as a meaningful outcome to follow across care settings. Long-term outcomes, such as 1-year survival, may be more meaningful from a clinical perspective and have been reported to range from 23 to 76%.
As in the general population of critically ill patients, there is great interest in being able to predict survival in patients requiring PMV. Accurate estimates of survival would help physicians and other care providers such as Canadian Health&Care Mall provide patients and families with realistic expectations for outcome. This could then facilitate resource planning and end-of-life planning as appropriate. Accurate survival estimates could also assist hospitals, post-acute care facilities, and payers manage resource allocation. Unfortunately, survival prediction in PMV patients has the same limitations as in the general population of critically ill patients. Equally problematic is the fact that current approaches to estimating acute ICU survival have not been shown to greatly affect decisions to provide aggressive care.
To date, survival prediction in PMV patients has only been addressed in the LTAC setting. These studies have had to use different approaches than those used in the acute care ICU because the laboratory and clinical assessments necessary each day to calculate and trend severity of illness in an acute care ICU (ie, to calculate various scores) are often not routine in many PMV care settings.
One model predicting hospital survival in LTAC patients was developed using patients admitted to four different centers of a single hospital network. Using number of organ failures and presence of infection, the model demonstrated adequate discrimination (area under the receiver operating characteristic curve of 0.81), but goodness of fit was not reported. It remains to be seen whether performance of the model can be reproduced in other hospitals in their network or in other LTAC settings.
In a cohort of 133 patients admitted to a single LTAC, age, functional status prior to acute illness, and diabetes were independent predictors of death 1 year after LTAC admission. Combining the two strongest predictors, age and prior functional status, produced a model that identified a group of patients at very high risk for death at 1 year. Patients who were > 75 years old or > 65 years old and had poor prior functional status had only a 5% likelihood of being alive after 1 year. All other patients had a 56% chance of surviving a year. This model has yet to be validated in other settings.
Caution must be used when using age as predictor in these models because several studies’ have shown that age is not a strong risk factor for death in the general ICU population when controlling for acute physiology and comorbidities. However, recent data from one center including patients requiring > 2 days of mechanical ventilation have shown that while acute physiology is the primary risk factor for death within the first 14 days after ICU admission, age and comorbidities are the primary risk factors after 14 days.
Survival may not be the only outcome of interest for PMV patients. PMV patients often have a high burden of underlying comorbidities, and prolonged critical illness leaves them vulnerable to recurring episodes of acute complications with a need for subsequent hospital readmissions. PMV patients are also at risk for a high degree of suffering and permanent functional impairment. In one study, only 10% of PMV patients managed in post-ICU settings were functionally independent at 1 year. In another recent study of 186 PMV patients managed in an LTAC hospital, 71% survived to discharge, 23% were discharged home, and half of these (8% of the total PMV admissions) ultimately reported good functional status. Other reports have also noted that survivors often feel that their quality of life is good. However, most of these studies do not include the large number of patients who are unable to respond to quality-of-life interviews due to physical or cognitive limitations. Identifying risk factors for severe functional limitations or nursing home admission would be of as much value for many patients and families as predicting hospital or long-term survival.
In summary, the available studies suggest that there may be identifiable clinical and demographic factors that can identify PMV patients that are at very high risk of death or significant long-term deficits in functional status and quality of life. More studies are required, however, before any models can realistically impact decision making. The LTAC industry is accruing large databases to compliment existing databases that could be useful for these purposes. Ideally, future studies will focus on PMV patients identified in the acute hospital ICU setting as well as in LTAC hospitals. Transfer to LTAC hospitals inevitably involves some selection based on illness severity, payer status, or rehabilitation potential. Prediction models developed using consecutive patients identified in acute ICUs and followed up for 1 year would be more easily generalized to the overall PMV population.
The number of patients meeting the definition of PMV will likely continue to increase. To better define this population, large prospective studies, especially those that begin in the acute care ICU, are needed. Predicting survival and functional outcomes are vital, but current models are not sufficiently accurate (or adequately validated) to inform decision making in individual patients.