This supplementary information is presented as submitted by the corresponding author. It has not been copy-edited by NTvG.
The aim of this research is to investigate if a classification method for readmissions, which is developed in England, can be used on Dutch administrative data to gain insight into the different types of readmissions. The understanding of the causal relationships and patterns of readmissions and prior admissions makes it possible to identify readmissions that may be the result of suboptimal care and are potentially preventable.
Retrospective analysis of Dutch administrative medical data.
A readmission is defined as an urgent, inpatient admission occurring within 30 days of discharge from a previous, inpatient admission in the same hospital, regardless of diagnosis or specialism. We assume that the urgent readmission is not planned on the basis of which the qualifications urgent and unplanned are used interchangeable. The used English categorisation is based on ICD-10 diagnostic codes and admission–readmission patterns. Administrative medical data of 18 hospitals is used. The 18 hospitals consist of 2 university hospitals, 9 teaching hospitals and 7 general hospitals, with which a representative picture of the situation in the Netherlands is assumed. Readmissions in 18 Dutch hospitals, occurring during 2009-2013, have been classified into seven categories: (A) complications; (B) anticipated but unpredictable hospital care; (C) preference; (D) artefact; (E) accident or coincidence related to different body system; (F) broadly related (related to the same body system); (G) other. In addition to the English classification, we have added category G, ‘other’, for admissions with ICD-10 Z-codes (socio-economic or psychosocial conditions) as main diagnosis. In England is the ICD-10 Z-codes as main diagnosis hardly used.
There were 120 343 readmissions, equivalent to 7,4% of hospital discharges. The readmissions were grouped into hierarchically exclusive categories: ‘Complications’ (17,7%), ‘Anticipated but unpredictable hospital care’ (10,3%), ‘Preference’ (0,1%), ‘Artefact’ (0,8%), ‘Accident or coincidence related to different body system’ (27,8%), ‘Broadly related’ (44,1%), ‘Other’ (4,6%).
The percentage of readmissions in the three types of hospitals varies slightly: university hospitals 5.8%, teaching hospitals 8.1% and general hospitals 6.6%. This pattern is possibly related to the differences in the proportion of urgent admissions: university hospitals 40.8%, teaching hospitals 57.5% and general hospitals 53.4%. University hospitals have relatively fewer readmissions in category B and G. A large proportion of readmissions in category G occur in teaching hospitals and general hospitals due to convalescence after a (surgical) treatment and readmission elated to obstetric care. The classification can vary greatly by specialty: pulmonary medicine, for example, has a relatively large number of readmissions in category B, whereas in surgery category A is relatively large.
The distribution of Dutch readmissions in the classification is different form the results of the English study. Category A and B contain fewer readmissions, while categories E and F are larger compared to the English results. In part, the differences can be explained because in England a more complete dataset is used that covers a larger period, making more readmissions identified.
The classification developed in England can be used on Dutch administrative medical data with a few adjustments. The classification provides insight into nature and occurrence of different types of readmissions. Advantage of this method is the use of existing administrative data, for which no additional medical registration is needed. Validation of the classification should reveal whether the classification is effective in the detection of suboptimal care and decreased patient safety. The classification gives hospitals insight into nature and cause of unplanned readmissions, so that targeted follow-up research of medical records and improvement actions are possible.