The goal of this analysis was to assess non-acute outpatient treated HK patients and chronic HK patients in terms of health-relevant outcomes. Further, the occurrence of hospitalizations, healthcare costs as well as dialysis initiation in relation to non-HK patients was explored. A retrospective design was applied using claims from the “Institut für angewandte Gesundheitsforschung Berlin” (InGef) research database. The InGef database consists of 6.7 million covered lives from about 60 different health insurances and includes the healthcare resource utilization and costs of services in an anonymized case-by-case individual format. For scientific research projects, an adjusted analysis sample of the InGef database has been drawn, including approximately 4 million covered lives which represent the German population in terms of age and gender (representing 4.8% of the German population and 5.6% of the German SHI population). Further, the analysis sample has been proven to have good external validity in terms of morbidity, mortality and drug use [16].
Patient identification
The complete study period included data from January 1st, 2014 through December 31st, 2016 with an enrollment period from January 1st, 2015 through December 31st, 2015. Selected patients were required to be observable within an individual pre-index period of 4 quarters and an individual post-index period including the index quarter and 3 consecutive quarters. The first observable diagnosis or prescription marked the index event/quarter.
Non-acute outpatient treated HK patients and those with chronic HK were identified and defined the main cohorts of interest. To that end, all patients who had at least one inpatient or outpatient ICD-10-GM (International Statistical Classification of Diseases and Related Health Problems, 10th revision, German Modification) diagnosis code of E87.5 “Hyperkalemia” or at least one prescription of polystyrene sulfonate (ATC code: V03AE01), were identified in 2015.. As a next step, we excluded all patients who were < 18 years in the quarter of their index event. Out of these patients, patients with non-acute outpatient treatment for hyperkalemia (cohort 1a) were identified by having at least two HK diagnoses and/or polystyrene sulfonate prescriptions with a restriction to one inpatient diagnosis and a follow-up outpatient diagnosis, indicating non-acute HK treatment.. Out of the cohort of non-acute outpatient treated HK patients, the subgroup of chronic HK patients (cohort 1b) was identified. These patients were required to have outpatient HK diagnoses or polystyrene sulfonate prescriptions in each quarter of the follow-up period. Figure 1 presents an overview of the steps that were performed to identify the respective HK patient cohorts as well as the respective cohort without HK from the database.
Matching approach
Individuals without hyperkalemia (cohort 2a/2b) were matched 1:1 to each identified hyperkalemia patient of cohort 1a and cohort 1b. Therefore, patients without any diagnosis of HK and without any prescription of polystyrene sulfonate during the whole study period (January 1st, 2014 to December 31st, 2016) were identified from the InGef database (cohort 2a/2b). Index quarters were assigned to adult patients without hyperkalemia based on the distribution of the index quarters in the cohorts 1a and 1b. An exact matching approach without replacement was applied for both matchings of the cohorts 1a and 2a and 1b and 2b, respectively. The matching parameters included age, gender, number of hospitalizations in the index quarter, prescriptions of RAASi in the index quarter (yes/no), and chronic kidney disease with related stages, taking the highest coded stage in the pre-index period into account.
The matching performance was evaluated by calculating absolute standardized differences of the cohorts as well as by statistical testing.
Outcome assessment
The assessment of patient characteristics included demographics in the index quarter and disease-specific comorbidities, such as CKD, HF, and diabetes mellitus type 2 (T2DM) in the pre-index period. Further, the mean updated Charlson Comorbidity Index (CCI) was assessed for each HK cohort (see Additional file 1).
As the individual number of hospitalizations in the index quarter were taken into account as a matching parameter, the all-cause and HK-related hospitalizations were analyzed in the 3 quarters following the index quarter.
Concomitant treatment was determined for RAASi therapy using ATC codes. Furthermore, use of sodium or calcium polystyrene sulfonate (SPS/CPS) stratified by disease-specific comorbidities within the HK patient cohorts, taking the most severe diagnosis for the respective comorbidity (CKD/HF) into account, was assessed. Prescriptions were analyzed in terms of mean number of prescriptions for SPS/CPS and proportion of heavy users (≥8 prescriptions in total or ≥ 2 prescriptions per quarter) as well as prescribing physician for SPS/CPS in the post-index period.
Healthcare costs stratified by different healthcare sectors, such as inpatient, outpatient, pharmaceuticals, and remedies and aids were assessed in the post-index period.
Time to dialysis initiation was assessed using the Kaplan-Meier method for all HK patients suffering from CKD (all stages) who did not have dialysis in the pre-index period. Therefore, the number of quarters between the first observable HK diagnosis and the occurrence of dialysis in the post index-up period was calculated. Patients were right censored in case of death, sickness fund switch or at the end of the observation period. Dialysis occurrence was assessed by OPS codes and defined EBM codes (see Additional file 1) at four possible timepoints (first quarter, second quarter, third quarter, fourth quarter) to analyze the time to event occurrence within the quarters of the follow-up period. Groups were compared with Log-Rank test.
Study outcomes were analyzed using counts and percentages and compared using the Chi-squared test for all categorical outcomes. Continuous outcomes were analyzed in terms of mean, standard deviation, 0.25 percentile, median, 0.75 percentile, minimum, maximum and sum and compared using the Wilcoxon signed rank test or the paired t test depending on the distribution of the differences.