Using Mpm Ii models

The computation of MPM probability is very straightforward. With the exception of the age variable, all variables have a 'yes' or 'no' response. The values for the variables listed for the particular model, the constant term b0, and the appropriate value(s) of b k noted for each variable with a 'yes' response, and the age variable obtained by multiplying the age by bk are entered into the following generic formula:

where bkxk is the value for the kth variable times the value of that variable (note that in MPM II the value of the variable is 1 if 'yes' and 0 if 'no', except for age which uses the actual age). The probability of mortality is obtained from the logit derived above using the following formula:

probability of mortality =

An example of a calculation of the probability of mortality of a patient at admission using MPM 0 follows.

A 60-year-old medical patient is admitted with cirrhosis and gastrointestinal bleeding. There is no history of renal insufficiency, cancer, or cerebral vascular incident. At the time of admission the patient is alert and oriented, and has a good color on room air; the heart rate is 140 beats/min and regular, and the blood pressure is 65/45 mmHg. There is no evidence of intercranial mass effect.

This short history has the following positive findings relevant to MPM 0: age, not a scheduled surgery patient, blood pressure 90 mmHg, history of cirrhosis, and an acute gastrointestinal bleed. Since we are performing an admission calculation, b 0 and the other appropriate values are extracted from Table 1 and inserted into the logit formula as follows:

lopr= -5,^6336 + 1.06427 + 1.13681 + 0,3%>3 +1.19098 + (0,03057 x 60) =0.15143.

This result is entered into the probability equation:

As with any ICU or similar model, the results from an MPM II calculation should be regarded as an estimate and a probability. Based upon a similar population of patients, it is expected that 54 per cent of those ICU patients with this constellation of variables will die during their hospital stay. The converse is that 46 per cent of such patients would be expected to survive.

Any multiple logistic regression probability model is developed from a specific population and can only purport to perform in that population. All the ICU models were developed from a diverse population of general medical and surgical patients in ICUs who were receiving intensive care and from data collected at specific times. Therefore their accuracy should only be assumed in similar settings. In an effort to deal with this, MPM II was designed to allow for customization to account for specific differences. For example, while a model might perform very well across the entire range of hospitals, there might be a benefit in customizing for a certain subgroup such as hospitals in a specific country. In addition, as the MPM II database grows it will be possible and necessary to customize the models to reflect general changes in mortality resulting from changes in treatment and improvements in the quality of care. Similarly, as the database grows, MPM II models for times beyond 72 h will be developed.

Lemeshow, S., Teres, D., Pastides, H., Avrunin, J.S., and Steingrub, J.S. (1985). A method for predicting survival and mortality of ICU patients using objectively derived weights. Critical Care Medicine, 13, 519-25.

Lemeshow, S., Teres, D., Klar, J., Avrunin, J.S., Gehlbach, S.H., and Rapoport, J. (1993). Mortality Probability Models (MPM II) based on an international cohort of intensive care unit patients. Journal of the American Medical Association, 270, 2478-86.

Lemeshow, S., et al. (1994). Mortality probability models for patients in the intensive care unit for 48 or 72 h: a prospective multicenter study. Critical Care Medicine, 22, 1351-8.

Was this article helpful?

0 0
Sleep Apnea

Sleep Apnea

Have You Been Told Over And Over Again That You Snore A Lot, But You Choose To Ignore It? Have you been experiencing lack of sleep at night and find yourself waking up in the wee hours of the morning to find yourself gasping for air?

Get My Free Ebook

Post a comment