¹Institute for Infectious Diseases and Infection Control, Jena University Hospital, Jena, Germany

Text

Introduction: Selecting the appropriate antibiotics treatment against moderate pneumonia is still a matter of debate. To develop a treatment rule for the decision of a beta-lactam monotherapy or a Macrolide beta-lactam combination therapy, we develop machine learning (ML) methods based on scoring models, trained on ex ante available etiological and clinical parameters. Propensity balancing is performed on each single treatment rule individually to make sure that for every rule the compliant patients (patients that were treated as the rule suggests) are balanced with the non-compliant patients (patients that were not treated as the rule suggests).

Methods: We consider characteristics of more than 10,000 hospitalised patients with moderate severity (non-intensive care unit patients) from the observational, prospective, multinational CAPNETZ study [1]. We start with finding a treatment rule based on one clinical variable. Then, treatment rules with two variables are tested, also in reasonable time. The number of scores containing more than two variables combinatorically explodes, but here we use a greedy-like algorithm to find the best combinations. The whole procedure is embedded into a 10x cross-validation scheme.

Results/conclusion: We observe that the performance of a treatment rule highly depends on the underlying balancing method. We addressed a similar research question in a recent study in which we searched for decision trees for the treatment rules [2]. However, using scores instead of decision trees enables to make the results more robust as can be seen from the cross-validation results and when validating with data from other studies. In line with [2], the first results show that macrolide combination therapy is beneficial for patients with high inflammatory indicators (CRP-value, leucocyte counts).

Members of the CAPNETZ study group are: M. Dreher, C. Cornelissen (Aachen); W. Knüppel (Bad Arolsen); D. Stolz (Basel); N. Suttorp, M. Witzenrath, P. Creutz, A. Mikolajewska (Berlin, Charité); T. Bauer, D. Krieger (Berlin); W. Pankow, D. Thiemig (Berlin-Neukölln); B. Hauptmeier, S. Ewig, D. Wehde (Bochum); M. Prediger, S. Schmager (Cottbus); M. Kolditz, B. Schulte-Hubbert, S. Langner (Dresden); W. Albrich (St Gallen); T. Welte, J. Freise, G. Barten, O. Arenas Toro, M. Nawrocki, J. Naim, M. Witte, W. Kröner, T. Illig, N. Klopp (Hannover); M. Kreuter, F. Herth, S. Hummler (Heidelberg); P. Ravn, A. Vestergaard-Jensen, G. Baunbaek-Knudsen (Hillerød); M. Pletz, C. Kroegel, J. Frosinski, J. Winning, B. Schleenvoigt ( Jena); K. Dalhoff, J. Rupp, R. Hörster, D. Drömann (Lübeck); G. Rohde, J. Drijkoningen, D. Braeken (Maastricht); H. Buschmann (Paderborn); T. Schaberg, I. Hering (Rotenburg/Wümme); M. Panning (Freiburg); M. Wallner (Ulm).

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.

Literatur

[1] Suttorp N, Welte T, Marre R, Stenger S, Pletz M, Rupp J, et al. CAPNETZ. The competence network for community-acquired pneumonia (CAP). Bundesgesundheitsblatt Gesundheitsforschung Gesundheitsschutz. 2016;59(4):475-81.
[2] König R, Cao X, Oswald M, Forstner C, Rohde G, Rupp J, Witzenrath M, Welte T, Kolditz M, Pletz M. Macrolide combination therapy for hospitalized CAP patients? An individualized approach supported by machine learning. European Respiratory Journal. 2019;54:11.