OBJECTIVE: This article presents methods for using administrative data to study multimorbidity in hospitalized individuals and indicates how the findings can be used to gain a deeper understanding of hospital multimorbidity. STUDY DESIGN AND SETTING: A Dutch nationwide hospital register (n=4,521,856) was used to calculate age- and sex-standardized observed/expected ratios of disease-pairing prevalences with corresponding confidence intervals. RESULTS: The strongest association was found for the combination between alcoholic liver and mental disorders due to alcohol abuse (observed/expected=39.2). Septicemia was found to cluster most frequently with other diseases. The consistency of the ratios over time depended on the number of observed cases. Furthermore, the ratios also depend on the length of the time frame considered. CONCLUSION: Using observed/expected ratios calculated from the administrative data set, we were able to (1) better quantify known morbidity pairings while also revealing hitherto unnoticed associations, (2) find out which pairings cluster most strongly, and (3) gain insight into which diseases cluster frequently with other diseases. Caveats with this method are finding spurious associations on the basis of too few observed cases and the dependency of the ratio magnitude on the length of the time frame observed.
OBJECTIVE: This article presents methods for using administrative data to study multimorbidity in hospitalized individuals and indicates how the findings can be used to gain a deeper understanding of hospital multimorbidity. STUDY DESIGN AND SETTING: A Dutch nationwide hospital register (n=4,521,856) was used to calculate age- and sex-standardized observed/expected ratios of disease-pairing prevalences with corresponding confidence intervals. RESULTS: The strongest association was found for the combination between alcoholic liver and mental disorders due to alcohol abuse (observed/expected=39.2). Septicemia was found to cluster most frequently with other diseases. The consistency of the ratios over time depended on the number of observed cases. Furthermore, the ratios also depend on the length of the time frame considered. CONCLUSION: Using observed/expected ratios calculated from the administrative data set, we were able to (1) better quantify known morbidity pairings while also revealing hitherto unnoticed associations, (2) find out which pairings cluster most strongly, and (3) gain insight into which diseases cluster frequently with other diseases. Caveats with this method are finding spurious associations on the basis of too few observed cases and the dependency of the ratio magnitude on the length of the time frame observed.
Authors: Rein Vos; Sil Aarts; Erik van Mulligen; Job Metsemakers; Martin P van Boxtel; Frans Verhey; Marjan van den Akker Journal: J Am Med Inform Assoc Date: 2013-06-17 Impact factor: 4.497
Authors: Josefien Buddeke; Michiel L Bots; Ineke van Dis; Frank Lj Visseren; Monika Hollander; François G Schellevis; Ilonca Vaartjes Journal: Br J Gen Pract Date: 2019-05-07 Impact factor: 5.386
Authors: Jeffrey H Silber; Joseph G Reiter; Paul R Rosenbaum; Qingyuan Zhao; Dylan S Small; Bijan A Niknam; Alexander S Hill; Lauren L Hochman; Rachel R Kelz; Lee A Fleisher Journal: Med Care Date: 2018-08 Impact factor: 2.983
Authors: Elisa Fabbri; Marco Zoli; Marta Gonzalez-Freire; Marcel E Salive; Stephanie A Studenski; Luigi Ferrucci Journal: J Am Med Dir Assoc Date: 2015-05-07 Impact factor: 4.669