Automated comparison of last hospital main diagnosis and underlying cause of death ICD10 codes, France, 2008–2009

Data sources

Definitions

The French definition of the main diagnosis has changed during the study period, from "condition that takes up the majority of resource use during the hospital stay" before march 2009, to "final diagnosis explaining hospital admission" after. However, this modification of definition had no impact on our results (results not shown).

In order to capture the pathology, which is the relevant information in our purpose, when the main diagnosis of the hospital database was a chapter XXI code (Factors influencing health status and contact with health services), the "main diagnosis" (MD) mentioned in this article was defined as the related diagnosis.

UCD is defined in volume 2 of ICD-10th revision as "(a) the disease or injury which initiated the train of morbid events leading directly to death, or (b) the circumstances of the accident or violence which produced the fatal injury".

Comparison method

Statistical analysis methods

In-hospital and out-of-hospital deaths were analyzed separately. The relationships between MD and UCD were studied according to age, gender, discharge-death time interval (in months) and main ICD Eurostat Shortlist chapters of UCD (corresponding ICD chapters): neoplasms (Chapter II), diseases of the nervous system and the sense organs (Chapters VI & VII), circulatory (Chapter IX), respiratory (Chapter X), and digestive (Chapter XI) systems, external causes of morbidity and mortality (Chapter XX), and one class for Others. Because they do not hold any information about the organs originally implied in the death process, imprecise UCDs (ICD-10 code in R99, R96.0, R57.9, R40.2, R09.2, I46.9, I99, I95.9, J96.0, J96.9, P28.5) were excluded from the comparison according to the category of UCD.

Univariate and multivariate log-binomial regression analysis [17] were used to study risk factors for independence vs. consistency, excluding non-informative cases. Relative risks (RRs) of independence and their 95% confidence intervals were estimated, crude and adjusted for age, gender, discharge-death time interval, length of stay, number of stays during the last year of life, and category of UCD. For each variable, the modal class was chosen as reference class.

Age, discharge-death time interval, length of stay, and number of stays during last year of life were included as continuous variables in order to perform trend tests.

Analyses were performed with SAS® version 9.3.

Place of death in/out hospital

Both independencies and non-informative UCD were higher for out-of-hospital than for in-hospital deaths: 19.5% vs. 8.5% and 7.6% vs.2.7% respectively.

Age and gender

On the whole, independence increases with age. However, considering the 15–34 age class, the proportion of independence was the lowest of all age classes for in-hospital deaths, and the highest for out-of-hospital deaths (results not shown).

UCDs are more often non-informative for deaths of age class 85 years and over than below 85 years: 4.8% vs. 2.1% in-hospital and 10.2% vs. 6.0% out-of-hospital. Non-informative UCDs were higher in females but this result was largely attenuated after adjustment for age.

Discharge-death time interval

Whereas the proportion of acceptable sequences remained roughly constant around 49%, similarities decreased progressively from 40.2% for in hospital deaths to 15.4% when death occurred more than 6 months after discharge, resulting in a decrease of consistency (acceptable sequence + similarity) from 88.8% to 61.6% (Figure 1).

As the discharge-death time interval increases, the proportions of both independencies and non-informative UCD increased. MD and UCD were independent in 8.5% of in-hospital deaths and 27.7% of out-of-hospital deaths occurring between six months and one year after discharge. The proportion of non-informative UCD reached 10.7% for deaths happening between six months and one year after discharge.

Cause of death

When the UCD was a neoplasm, consistencies reached 93.7% and the MD of last hospital stay was more often similar to UCD (54.9%), much higher than for any other cause of death (Figure 2). Consistencies were around 83% for the other categories of UCD, similarities varying from 17.9% for neuro-sensorial diseases to 37.7% for digestive diseases.

By definition, for affections due to external causes, MD should belong to ICD-10 Chapter XIX [13] and UCD should belong to chapter XX [11]. They can therefore never be similar, but such cases were detected as acceptable sequences by our algorithm.

Independence vs. consistency

Considering out-of-hospital deaths, independence represented 21.1% of cases. The proportion was especially high for deaths in the 15–34 years class, or for death with an external cause. It was positively associated with the discharge-death time interval (ptrend < .0001) and negatively with length of stay (ptrend < .0001) and number of hospitalizations in the last year of life (ptrend < .0001). After adjustment, the strength of these associations weakened, but the associations with discharge-death time interval and with category of UCD other than neoplasms remained noticeable. Age class 15–34 years was no longer associated with an increased risk of independence.

Principal findings

We proposed an automatic method of comparison of the main diagnosis (MD) of the last hospital stay to the underlying cause of death (UCD) relying on Iris software, in order to determine their consistency or independence. This method proved able to analyze automatically 91.7% of the 421,460 submitted deaths having occurred in France in 2008–2009 within one year from last discharge. The main reasons for rejects were MD not accepted as valid causes of death and iatrogenicity.

In most cases, MD of last hospital stay and UCD were consistent, or in other words, referred to a same train of events leading to death: 88.8% of in-hospital death and 72.9% of deaths occurring after discharge.

The distribution of consistency and independence according to socio-demographic and medico-administrative variables gave expectable results: independence was more frequent in elder patients, likely because they suffer from multi-pathologies, or as the discharge-death time interval grew (8.5% of in-hospital deaths, 14.3% when death occurred within one month after discharge and 27.7% within 6 to 12 months), or for non-neoplasms UCDs, which had already been noted in former studies [5, 6, 8].

A long last stay or numerous hospitalizations in last year of life were associated with higher independence for people dying in hospital but with lower independence for people dying out-of-hospital, which may seem a paradox. However, independence was still lower for patients deceased in hospital after a very long stay (11.4%) than for patients deceased after discharge of a very short stay (17.9%). Possibly, these are markers of severe and complex medical situations that more often involve multi-pathologies. One hypothesis behind these results would be that complexity explains the result for in-hospital deaths, but in the same time, severity is associated with a greater probability for the physician certifying the death to recall the hospital main diagnosis for out-of-hospital deaths.

Besides, this study has shown that MD should not be used as a proxy of the UCD, even for patients deceased in hospital, since MD and UCD are similar in only 40% of in-hospital deaths. This result accounts for the difference of definition and coding context of MD and UCD.

Results in relation to other studies

The levels of consistency found in France are similar to those previously measured in Sweden: 89% for in-hospital deaths in both countries and 71% in France vs. 68% in Sweden for out-of-hospital deaths [7]. In the Swedish study, hospital case summaries for some hospital deaths were investigated showing that inconsistency between MD and UCD was often due to certification errors. Among non-consistent cases, our algorithm was designed to distinguish independency and non-informative death certificates. However, only a look back on a series of medical records will assess the rate of coding errors on MD or certification errors on UCD leading to misclassification.

Strengths

We have proposed a formal definition of the concept of independence between the main condition treated during a hospital stay and a subsequent death.

Using Iris Software, we have designed a language-independent method of comparison of ICD-10 codes of MD and UCD, which is based on international standards. The use of international procedures and tools guarantees that the method can be reproduced by any country. This widely automated method makes the comparison feasible on very large national datasets.

Limits

The appraisal of independence could likely be improved. Generally speaking, independence is probably under-estimated because the MMDS knowledge table through which the causal relation are judged was designed to appraise causality between two causes, knowing that a medical doctor had declared them linked. They therefore accept "possible causality". A way to limit this bias would be to build a stricter table, aiming at only capturing "probable causality".

Alternatively, in some cases, independence is likely to be over-estimated. An example is the recording of two different primary neoplasms as MD and UCD, resulting in an "independent" label, whereas one of the codes is probably erroneous. Indeed, in some frequent cases, both codes most likely refer to the same pathology (eg: malignant neoplasm of colon (C18) and malignant neoplasm of rectum (C20)). More generally, users' feedback might bring a few improvements of the algorithm and reduce these issues.

Another source of independence over-estimation lies in considering only the hospitalization main diagnosis. For patients suffering from multiple pathologies, this may lead to the labeling "independent", whereas a pathology similar or causally related to the UCD was in fact taken care of during the hospital stay. Comparing all the conditions mentioned in the hospital discharge abstract to the UCD would resolve this; it would need the development of a more complex algorithm.

Future research on hospital quality of care assessment

We believe that the concept of independence between MD and UCD, along with the practical method of appraisal exposed herein, could be useful for the construction of post-hospital mortality indicators. Mortality is currently used in several countries to compare hospitals quality of care, although this has been criticized [19–21]. For example, hospital standardized mortality ratios (HSMRs) are used in an increasing number of countries including England, Canada and the United States [22–24]. Relying on hospital administrative data only, these indicators give an overall measure of in-hospital mortality, adjusted for available case mix factors. Nevertheless, factors such as length-of-stay and transfer or discharge patterns, which vary between hospitals, affect in-patient death rates [25–28]. Therefore, taking into account the deaths occurring out of hospital and using time-based indicators such as total 30-days from admission mortality is necessary and has a significant impact on HSMRs [23]. On the other hand, mortality after discharge is linked to several other factors and may not reflect actual hospital performance [29–33]. Indeed, our results confirm that the longer the time after discharge, the higher the probability for an independent cause of death to occur. The exclusion of deaths independent of the MD from the deaths allocated to the hospital might thus improve the accuracy of potential mortality-based quality-of-care indicators.

The method exposed in this paper is general and potentially applies to all MD/UCD configurations. However, for some specific issues, further developments are necessary. This is the case for iatrogenicity which is systematically rejected by Iris. Although these cases are rare (3.2%) and likely do not influence much HSMR estimates, they are informative for quality of care evaluations, and should therefore be studied following a different methodology.