Data collection
All data were collected by a researcher (MS) from patients’ electronic care records (ECRs) between 28th July 2016 and 4th October 2016, as close to the patients’ clinic visits as possible. Demographic and disease related data included age, gender, number of comorbidities, forced expiratory volume in 1 s (FEV
1; percentage predicted) and latest sputum microbiology. The Charlson Comorbidity Index (CCI), a clinical risk assessment tool, was used to quantify the burden of certain comorbidities and predict mortality, with a higher score indicating a greater burden and greater risk of mortality [
12]. A newer, validated bronchiectasis-specific comorbidity index, the Bronchiectasis Aetiology and Comorbidity Index (BACI) was also calculated [
13].
Data on medicines regularly prescribed for patients by their general practitioner (GP), including drug name, prescribed dose and frequency of administration, quantity prescribed and date the drug was first and last issued, were collected. This information was used to compile a list of all medicines each patient was prescribed at the time of their clinic visit. Different strengths and formulations of the same medicine were counted separately.
The total number of prescribed medicines was determined by absolute count. Patients’ medication counts were used to identify whether or not three thresholds for polypharmacy (‘≥ 4 medicines’, ‘≥ 10 medicines’’ and ‘≥ 15 medicines’) had been exceeded.
MRCI scores were calculated using information collected about prescribed medicines [
8]. The MRCI consists of 65 items divided into three sections: ‘Dosage forms’, ‘Dosage frequencies’ and ‘Additional instructions’. Each item is weighted according to the relative degree of complexity it contributes to the regimen [
8]. Scores begin at 0 and because patients may be on any number of medicines, there is no maximum MRCI score. There are no widely recognised thresholds that represent particularly low, medium or high MRCI scores.
The primary outcomes of interest were oral antibiotic usage for suspected pulmonary exacerbations in the 6 months prior to clinic attendance, and IV antibiotic usage in secondary care in the 2 years prior to clinic attendance. Secondary outcomes included the total number of days’ treatment with IV antibiotics and admissions to hospital (all-cause and bronchiectasis-related) in the 2 years prior to clinic visit.
Medications prescribed for patients on an acute basis by GPs, including short courses of antibiotics, are recorded in patients’ ECRs. The following information regarding oral antibiotics prescribed for patients was recorded: issue date, antibiotic name and dose, frequency of administration, and duration prescribed (days). Indications for antibiotics are rarely specified on prescriptions issued in primary care. Oral antibiotic use for a pulmonary exacerbation was suspected if one or more of the following factors were present: the patient was prescribed an appropriate antibiotic (e.g., doxycycline, amoxicillin, co-amoxiclav, clarithromycin) for 14-days, and/or a sputum sample was collected and tested within a week of the prescription being issued [
11]. Clinic notes were also consulted to check if the patient had reported any exacerbations since their last visit. In cases of ambiguity, the patients’ consultant physician provided clinical judgment of whether, or not, an antibiotic had been prescribed for a suspected pulmonary exacerbation. Only medicines prescribed in the past 6-months are available to view at the time of data collection, hence the 6-months time period for documenting oral antibiotic use.
IV antibiotic use was determined using data from patients’ discharge summary letters, which are available to view on the ECR and records extend back several years. Data regarding all IV antibiotics (date prescribed, name of antibiotic and duration of therapy) prescribed for patients for the treatment of an acute exacerbation of bronchiectasis and all hospitalisations in the past 2 years were collected. There was less ambiguity regarding the indication of IV antibiotics during hospital admissions because the indication of prescribed treatment during a hospital stay is routinely documented. IV antibiotic use and secondary outcomes were investigated over a longer period than oral antibiotic use to mitigate for seasonal variations in exacerbation rates.
A power calculation, based on a comparison of the mean number of times an oral antibiotic was prescribed for a suspected exacerbation in the preceding 6 months between patients who were prescribed 10 or more medicines and those who were prescribed less than 10 medicines, was performed prior to data collection. McCullough et al. [
14] suggested a difference of two exacerbations within a 12-month period was importance clinically important difference. Assuming the standard deviation of exacerbations in 6 months was 1 (based upon findings from McCullough et al. [
14]), with 100 patients, the power to detect a difference of one exacerbation within 6 months as significant at the 5% level was over 80%.
Data analysis
All data were entered into IBM SPSS (Version 23). The Mann–Whitney U test was used to investigate if there was a significant difference in the primary and secondary outcomes of patients below and above the three thresholds for polypharmacy. Effect sizes (r) were calculated using the calculation ‘r = z/square root of N, where z is the standardized test statistic and N = number of cases’ [
15].
Correlations between MRCI and the primary and secondary outcomes were investigated using Spearman rank-order correlation coefficient (Spearman’s correlation,
rs). Correlations were considered strong/large if
r = 0.5–1.0; moderate/medium if
r = 0.3–0.5; weak/small if
r = 0.1–0.3 [
15].
To further investigate the association between polypharmacy and IV antibiotic use, patients were grouped into those who had had a course of IV antibiotics for the treatment of an exacerbation in the past 2 years and those who had not (status: did not receive IV antibiotics = 0, did receive antibiotics = 1). Chi square (χ2) tests were conducted to determine whether, or not, there was a difference in receipt of an IV antibiotic in the past 2 years between people who were above and below each of the three polypharmacy thresholds. Where significant differences were identified, unadjusted odds ratios were calculated. Other demographic and disease-related factors (age, gender, testing positive for P. aeruginosa or H. influenzae), and comorbidities may contribute to a higher exacerbation rate. Logistic regression was then used to adjust odds ratios for any other factors that could have potentially influenced the need for IV antibiotic use. The significance level was set at p ≤ 0.05 for all statistical tests. The associations between oral antibiotic use and polypharmacy were not investigated in this way because there was less certainty over the indication of oral antibiotic use.