At Erasmus MC we constantly work to analyze the outcomes of health and healthcare so that we can use these outcomes to improve our quality. We publish quality results and outcome figures that give a good indication of the quality of our care: ‘Our results’.
HSMR and SMR death data
Every year, the Dutch Healthcare Authority (NZa) asks all hospitals in the Netherlands for an average mortality rate for the entire hospital, which is known as the ‘Hospital Standardized Mortality Ratio’ (HSMR), an adjusted hospital mortality figure. This ratio is an indicator of the number of patient deaths in a hospital, adjusted for the expected mortality, compared to the national average of hospital deaths in a given year. Those in favor of this ratio see it as a measure of the quality of the healthcare provided by a hospital. In addition, hospitals must also publish their SMR (Standardized Mortality Ratio) death data. SMR is the ratio of the observed number of deaths in a hospital to the expected number of deaths, by diagnosis and patient groups. To calculate the HSMR, only clinical admissions that fall within one of the diagnosis groups are included. The HSMR is the average of all SMRs.
Objecting to the HSMR
A higher than average HSMR result may indicate possible shortcomings in the quality of healthcare provided. However, the current method does not take into account the seriously ill patients, often with highly complex conditions, admitted to a university medical center (UMC), and it compares the less sick patients with the same diagnosis admitted to a general hospital. Furthermore, only deaths per hospital admission are included, and not deaths that occur outside the hospital. For some diseases, multiple admissions are required. Mortality is these cases therefore cannot be measured by a single admission. In addition, the admission and discharge policy for seriously ill patients differs among Dutch hospitals, partly depending on the availability of palliative care and hospices, making the figures incomparable. This is expressed in Erasmus MC publications (see references below).
Mortality in 2017
In 2017, 35,343 patients were admitted to Erasmus MC, of whom 864 (2.4%) died.
The mortality rate for 2017 was calculated based on the 2017 national calculation model. An HSMR of 100 means that the expected mortality is equal to the observed mortality. A ratio of less than 100 indicates that a hospital’s mortality rate is lower than expected; a ratio greater than 100 indicates that a hospital’s mortality rate is higher than expected. Erasmus MC’s HSMR for 2017 is 121; the 95% confidence interval is 113 - 130. This HSMR is significantly higher than the national average and higher than the HSMR of other UMCs.
SMRs in 2017
Our SMR figures are shown from page 32 onwards in the PDF document at the bottom of this page. These figures express the mortality rates in the various diagnosis groups and underlying patient groups. In 2017, some SMRs are significantly different statistically from the national average. In a number of these cases, this is a structural deviation from the national average that can be attributed to factors mentioned below.
For example, the high SMR for conditions that occur during the perinatal period can be explained by the fact that Erasmus MC Sophia Children’s Hospital has a large neonatal intensive care unit (NCIU) with a relatively large number of extremely premature babies. This means that the ratio will deviate from the national average, as will be the case at other NCIUs. Moreover, Erasmus MC Sophia Children’s Hospital is a center for neonatal ECMO (extracorporeal membrane oxygenation), as is Radboudumc.
The fact that Erasmus MC repeatedly has an SMR that deviates significantly in the ‘injury and poisoning’ cluster is probably due to the fact that Erasmus MC is a trauma center, which means that the most seriously injured trauma patients from the region are treated here. The current system used for the SMR does not specifically adjust for the seriousness of the injury. In the current model, a patient admitted with a severe coma will have the same calculated risk as someone who is not comatose.
The consistently high SMR in the ‘respiratory diseases’ diagnostic group is probably due to the fact that many severely weakened patients with other underlying conditions are treated at Erasmus MC, who ultimately die of a respiratory disorder.
It is more difficult to clarify why several diagnosis/patient groups (namely infectious and parasitic diseases, cardiovascular diseases, and diseases of the digestive system) have a significantly higher SMR for the first time.
On the basis of the final figures for 2016 and 2017, a study was started into the possible underlying causes for the figures, with the aim of improving the quality of the care provided.
Two findings have emerged so far.
The information used to calculate the HSMR, longer than expected LOS, and readmission ratio is collected by medical coders, using information provided in discharge letters. However, it appears that discharge letters often contain insufficient information to ensure accurate coding. To facilitate a complete transfer of the clinical course to medical coders, a standardized discharge letter is under development. This standardized letter will be included in the electronic patient record (HIX).
A model adjusting for distorting factors, such as the severity of the diagnosis on admission, is used to calculate the (H)SMR. There are indications that the coding used (ICD-10) may not differentiate enough to correspond to the actual severity of the diagnosis. This can lead to unjustifiably low mortality risks and therefore to a higher HSMR than is actually the case. This is currently being investigated further.
1. Pouw ME, Peelen LM, Moons KG, Kalkman CJ, Lingsma HF. Including post-discharge mortality in calculation of hospital standardised mortality ratios: retrospective analysis of hospital episode statistics. BMJ 2013; 347: f5913.
2. Gestel YR van, Lemmens VE, Lingsma HF, de H, I, Rutten HJ, Coebergh JW. The hospital standardized mortality ratio fallacy: a narrative review. Med Care 2012; 50(8): 662-667.