Classification of hospital pathways in the management of cancer: application to lung cancer in the region of burgundy

CONTEXT

The evaluation of national cancer plans is an important aspect of their implementation. For this evaluation, the principal actors in the field (doctors, nurses, etc.) as well as decision-makers must have access to information that is reliable, synthetic and easy to interpret, and which reflects the implementation process in the field. We propose here a methodology to make this type of information available in the context of reducing inequalities with regard to access to healthcare for patients with lung cancer in the region of Burgundy.

METHODS

We used the national medico-administrative DRG-type database, which gathers together all hospital stays. By using this database, it was possible to identify and reconstruct the care management history of these patients. That is, by linking together all attended hospitals, sorted chronologically. Eligible patients were at least 18 years old, whatever the gender and had undergone surgery for their lung cancer. They had to be residents of Burgundy at the time of the first operation between 2006 and 2008. Patient's pathway was defined as the sequence of all attended hospitals (hospital stays) during the year of follow up linked together using an anonymised patient identifier. We then constructed a pathway typology of pathway using an unsupervised clustering method, and conducted a spatial analysis of this typology.

RESULTS

Between 2006 and 2008, we selected 495 patients in the 4 administrative departments of the Burgundy region. They accounted for a total of 3821 stays during the year of follow-up. There were 393 men (79%) and the mean age was 64 (95% confidence interval: 63-65) years. We reconstructed 94 pathways (about five per patient). Here, neighbourhood's cares accounted for 41% of them, while 44% included a surgical intervention outside the region of Burgundy. We constructed a pathway typology with five classes. Spatial analysis showed that the vast majority of initial surgeries took place in the major regional centres.

CONCLUSION

The construction of a pathway typology leads to better understanding of the reasoning that lies behind the movements of patients. It opens the way for analysis of the collaboration between the different healthcares establishments attended, which should bring to light associations that need to be developed.

Testing for course patterns in Crohn's disease using clustering analysis

Using clustering analysis, we sought to identify groups of patients on the basis of the disease course among a population of 177 patients with Crohn's disease and followed for 3 years or more, starting from the first frank exacerbation of the disease. The first 36 values of a monthly clinical score represented the active variables of the clustering analysis. This method yielded 2 course groups, A and B, of 95 and 82 patients respectively. The unfavorable course in group A was characterized by the persistence of the clinically active disease at 3 years, whereas group B individuals achieved complete clinical remission within 2 years of onset on the average. Among the initially known clinical data which could explain the course, only the incidence of an occlusive syndrome was higher in group B, which showed a more favorable course. Although we applied clustering analysis to a patient sample over a period of only 3 years, our results do suggest the existence or 2 primary course groups within the population of patients with Crohn's disease. It would appear that the disease course cannot be predicted from the clinical parameters present at the time of onset, but rather becomes apparent during the course of the first 2 years.