Dear Colleague,

First of all you have to collect at least 200 patients. Otherwise, the result does not make sense. However, to obtain the SMR according to your example, you have to follow the following steps:

1) You have to transform the percentages in numbers.

In your example you will have: 0.1, 0.17, 0.32, 0.06, 0.17

2) You have to sum the numbers: the total is 0.82

This is the number of expected deaths

3) The formula to obtain SMR is: observed hospital deaths/predicted hospital deaths. In you example SMR is 1/0.82= 0.21

See also Rhodes et al Prospectively defined indicators to improve the safety and quality of care for critically ill patients: a report from the Task Force on Safety and Quality of the European Society of Intensive Care Medicine (ESICM). Intensive Care Med (2012) 38:598–605 DOI 10.1007/s00134-011-2462-3

Best regards

Maurizia Capuzzo

Thank you so much Mrs Capuzzo..

Dileep Unnikrishnan

Please find the probable explanation based on our experience with analysis of PIM scores.

1. Taking your example- the percentage predicted is 10%, 17%, 32%, 6% and 17% for patient 1, 2, 3, 4, 5. The mean probability of death for these 5 patients is 10+17+32+6+17= 82%/5 patients = 16.4

2. The observed mortality is 1/5 = 20%

3. The SMR= Observed /predicted = 20/16.4= 1.21 which indicates the observed is higher than expected or predicted.

4. When we run statical tests (say for 200 patients) and divide the entire sample into 10 centiles of say 20 patients each starting from the lowest probability to 100% probability; the mean for each centile of patients or the group of 20 patients is given by the statistical test. In other words the statistical test will give you the mean for 20, 20 ,20,...patients and so on for the 10 groups that you want.

5. This is how you can get the mean for each group of 20 patients or 10 patients depending on your sample size of 200 or 100.

6. You will get the mean of these 10 groups which will invariably be lower than the observed deaths in most units and therefore the SMR will be more than one.

Hope this helps.

Thanks

Jhuma

Thank you madam..

Is there a software which can help calculate the SMR?

Hello mam Maurizia Capuzzo I was searching same question asked by Dileep Unnikrishnan I found it will be helpful to me as I am calculating SMR for ICU, but I m confused with Do I have to calculate the APACHE score for all admissions(discharge+expired) in ICU per month (specified time period)? if we have 100 admissions in ICU per month ...

then the percentage into numbers and sum up that numbers.

Please guide me for predicted no. of mortality , how to consider patients? all admissions per month and their individual APACHE predicted mortality ? or we have to calculate only APACHE predicted mortality of only expired patients per month?

Thanks

Dear Iva,

Generally SMR is calculated on a yearly basis (for instance, from 01/01/2020 to 31/12/2020).

You have to include all the patients admitted to ICU in the period (year). It does not make sense to consider only dead patients.

If you have the number of the admissions in the period (as usually recorded by the hospital administration), you have to remove the readmissions (we include only the first ICU admission for the patients having two or more ICU admissions in the period considered). This allows you to have the number of the patients admitted, which is the information required for the computation of the observed mortality rate. If in your country there are rules for limitation of care, you should move from the number of patients ICU admitted the number of those admitted to ICU with limitations of care (for example, those patients who have previously declared that they do not want to be intubated).

I tell you the practical information about how we manage data in the ICU where I worked.

In the early afternoon (or at the time of the day we are less busy) we collect the clinical data for each patient admitted in the previous 24h and we report the values of severity score variables in a database. Usually, we update the database daily adding any information we are interested in the suitable fields (intubation, tracheostomy, vasoactive amines etc.). At the time of discharge from the ICU, we record the date of ICU discharge and whether the patient has been discharged alive or dead.

I know that other ICUs collect general clinical data, information about the clinical course, and ICU discharge information at the time of ICU discharge or in the subsequent days. Personally, I think that collecting data during the patient stay in ICU is better because it may reduce the risk of forgetting any information, but you can do it as you prefer.

Every six months (usually in September and February) we ask the hospital administration to give us the date of the hospital discharge (and the vital status) of all the patients admitted either between the period January-August, or between the period July-February. For each patient, we record this information in the database.

You have to calculate APACHE II for each patient admitted to ICU in the period (year)

For each patient, you calculate the sum of the physiological score (12 variables) + age points + Chronic Organ Insufficiency points. Then you identify the Diagnostic category and use the logit to calculate the mortality predicted by APACHE II for each patient. You can see the original paper (Knaus WA, Draper EA, Wagner DP, Zimmerman JE (1985). "APACHE II: a severity of disease classification system". Critical Care Medicine. 13 (10): 818–29. doi:10.1097/00003246-198510000-00009. PMID 3928249. ) or https://sfar.org/scores2/apache22.php

The SMR is the ratio between observed mortality ratio (no. of patients admitted to the ICU that die in the hospital/number of patients admitted to the ICU per year) and expected hospital mortality (arithmetic sum of the individual probabilities of death for each patient admitted to the ICU according to the severity score/no. of patients admitted to the ICU).

The SMR allows the comparison of the results of the ICU with those predicted by the model. Moreover, it can allow the comparison of your ICU over time (different years), and the comparison of your ICU with other ICUs.

I am not sure I understand all the questions, but I hope these answers are helpful to you.

Maurizia Capuzzo

Thank u so much mam Maurizia Capuzzo . The information shared by u will be helpful and well described.

The only problem with getting the predicted mortality is that most of us use APACHE 2 as the admission score and the predicted mortality rate is from a very old study.

I agree with Anirban. Nevertheless, an old scoring system is better that no scoring system...