The study uses a retrospective, matched cohort design, and has been conducted in accordance with applicable subject privacy requirements and the guiding principles of the Declaration of Helsinki of 1964. The analysis was based on secondary claims data and, as such, the consultation of an ethics committee was not required [10].

Analyses were based on anonymized routinely collected claims data from the German Statutory Health Insurance (SHI), covering the period of 2015–2019. The dataset comprises information on up to five million people insured at 19 sickness funds, representing 6.3% of the SHI population in Germany. The data is provided by GWQ ServicePlus AG, a joint venture of medium-sized health insurers in Germany. Overall, 87% of the German population is insured within the SHI system [11]. Based on comparison with official statistics published by the Germany Health Ministry on the German statutory health insurance (“KM6”), the dataset is representative of the German SHI population in terms of age and gender distribution [12]. Diagnostic data include all diagnoses documented during physician outpatient contacts and patient hospital stays. Laboratory or clinical parameters were not included. A general description of the claims database in the German setting can be found in Swart et al. [10].

The initial dataset contained all adults aged 18 years and older in the years 2015–2019. Individual information on the year of birth is aggregated to 5-year intervals in the anonymization process. To ensure that all patients are at least 18 years of age, the age was computed as the difference between the year under study and the upper bound of the age interval.

Based on the literature and in consultation with a medical expert panel, the following UCs were defined as potential risks: asthma, chronic obstructive pulmonary disease (COPD), osteoporosis, rheumatoid arthritis, depression, immunodeficiency, heart failure (HF), chronic heart disease (CHD), chronic kidney disease (CKD), diabetes mellitus type 2 (DMT2), and diabetes mellitus type 1 (DMT1). While there may be other UCs with an impact on pertussis, in this manuscript UCs always refer to these risk constellations. For a detailed description of how these diagnoses were validated with the corresponding International Classification of Diseases (ICD), 10th revision, German Modification (ICD-10-GM) diagnosis codes and prescription codes (based on the Anatomical Therapeutic Chemical [ATC] Classification), see Table S1.

An individual was identified as having a specific UC if either an inpatient or an assured outpatient diagnosis was documented in at least two out of four quarters of two consecutive years. Patients were included in the study population if they were diagnosed with the respective UC in the years from 2016 to 2019. As the year under consideration always had to be validated by the previous year, data from 2015 was used as a wash-in period. A patient may belong to more than one UC group. A matched cohort design was adopted; patients with at least one UC were matched to controls without UC by exact matching. Matching variables also included age in 5-year intervals and sex of the patients. Matched pairs were observed for the same period of time (until one of them died or left the SHI).

Pertussis was broadly defined as having at least one diagnosis of ICD-10-GM A37.0 (Whooping cough due to Bordetella pertussis) or A37.9 (Whooping cough, unspecified). The diagnosis could either be in- or outpatient. As a sensitivity analysis, a narrow definition only considering the diagnosis of ICD-10 A37.0 is reported. Incident pertussis cases were defined as cases without a pertussis diagnosis in the preceding year; hence, information from 2015 was again used as washout period. The index quarter was defined as the quarter when pertussis was first observed.

A pertussis-associated complication was counted if it occurred in the index quarter and was not observed in the four quarters preceding the index quarter and, hence, was incident (except for hospital visits). We distinguished between severe and less severe complications. Severe complications comprised pneumonia, rib fractures, and all-cause hospitalizations but not including the following less severe complications. Otitis media, encephalopathy, abnormalities of breathing, seizures, inguinal and umbilical hernia, intracranial hemorrhage, incontinence, loss of weight, sarcopenia, and gait abnormality were counted as less severe complications (see Table S2 in the supplementary material for used ICD-10-GM codes). Results were also summarized as any complications that occurred. In the sensitivity analyses, the complication could also be diagnosed in the quarter prior to the index quarter and/or in the quarter following the index quarter. In the following, data based on the broad pertussis definition are presented unless described otherwise.

In the analysis of pertussis incidence, rates were calculated for both the narrow and the broad definitions of pertussis and reported separately for patients with any UC, with no UC, and with one of the specific UCs, as well as for age groups (i) 18–49, (ii) 50–59, and (iii) 60 and older.

To analyze the risk of pertussis for each UC, logistic regression models, controlled for age and sex, were fitted to the pooled data of the matched population (cases and controls) from 2016 to 2019. Separate models were fitted for patients with any UC, and for each UC compared to matched persons without UC respectively. In a sensitivity analysis, a model with all UCs as covariates was fitted to check the robustness of the results. As a result of small sample sizes for the narrow case definition, a model with a single coefficient for any UC was fitted, not considering each UC separately. The resulting odds ratios are reported with corresponding 95% confidence intervals (CIs).

In the analysis of pertussis-associated complications, patients with pertussis and UCs were compared to patients with pertussis and no UC. A complication was considered to be pertussis-associated, if it was observed in the same quarter as the pertussis. Complications were counted for all years under study and reported as complication rates in patients with pertussis and with UC and without UC. To estimate the incidence rate of complications, a negative binomial regression model controlled for age, sex, and any UCs was fitted to the number of individual complications in the pertussis cohort. The analyses for complications were rerun in sensitivity analyses, where complications could also be observed in the quarter before or after the pertussis quarter.

To assess whether there is an association between specific age groups and the risk of complications, a negative binomial regression model, including age groups 18–49, 50–59, and 60+ as covariates, was estimated.

All analyses were performed with R (version 4.1.3) using the MatchIt package for matching, the glm function for logistic regression models, and the glm.nb function of the MASS package for negative binomial regression models.