Descriptive characteristics of the CAB measure, steps involved in its implementation, and the measure output were evaluated against the same parameters for existing measures of cumulative anticholinergic exposure to determine their suitability for application in administrative claims analyses. Data from a large US health dataset were used in the evaluation.
Descriptive Comparison of the Measures
The following details of each measure have been summarized: data requirements for implementation; whether it considers anticholinergic cumulative exposure and potency as well as dose; the disease context for which the measure was developed; whether its development was based on a specific anticholinergic scale; its flexibility for use with all anticholinergic scales; its mathematical properties (e.g., average vs. aggregated score; score boundaries); the definition of exposure period; and considerations for the categorization of the resulting scores (Table 1). Other potential measures were also considered initially, including the Drug Burden Index with a modification to extrapolate anticholinergic burden longitudinally [20, 21]. However, it was not included here as it does not consider anticholinergic potency and would therefore likely have similar limitations as those of the cumulative dose measure [17]. Additional details on the characteristics of anticholinergic measures assessed can be found in this published review [18].
Table 1 Features of anticholinergic burden measures Quantitative Evaluation of the Measures
The three measures were applied in a claims database to determine their suitability for use in retrospective studies relying on administrative data, and to evaluate correlations between the measures.
Data Source
Data from the US-based Truven MarketScan databases from 2012 to 2015 were used. These databases contain nationally representative healthcare data on patients insured commercially, or with employer-sponsored supplemental Medicare coverage. Data include demographics, diagnoses, records of outpatient and inpatient medical services, and pharmacy claims for over 84 million people linked at the individual level.
Cohort Identification
A 2.5% random sample was selected from individuals in the Truven MarketScan dataset aged ≥ 65 years between January 1, 2012 and December 31, 2014. The sample was computer-selected using a random number generator where all individuals aged ≥ 65 years had an equal likelihood of being selected. Cohort members were enrolled on the first visit during the identification period at which an individual was ≥ 65 years old (index date). A minimum of 12 months of pharmaceutical prescription data were required after enrolment to ensure that anticholinergic exposure could be estimated consistently across all cohort members.
Scale Selection
Although several anticholinergic scales with a hierarchical anticholinergic weighting system exist, the ACB scale was selected for the implementation of the three evaluated measures as it is a validated scale with a wide coverage of medications with anticholinergic effects and the inclusion of numerous high-potency medications (see Online Resource Figure 2) [22]. Although the ACB scale has been used in other jurisdictions [23,24,25], as it was developed in the US it was expected to have better coverage of medications included within this dataset than those developed outside the US [18].
Data Evaluation for Measure Implementation
A descriptive evaluation of the data and steps required for implementation of the three measures was conducted. Each measure was evaluated based on the extent of data cleaning and merging required, the ease of extraction and the amount of matching of medication information with the US Food and Drug Administration’s National Drug Codes (NDC).
Various degrees of data cleaning were required to implement each measure, and to merge medication names, routes of administration, tablet strength and unit of the active ingredient. Medication characteristics missing from the dataset were extracted from the NDC codes available in MarketScan [26].
Measurement of Cumulative Anticholinergic Exposure
Cumulative anticholinergic exposure was calculated for a 1-year period per cohort member, according to the average daily dose, cumulative dose and CAB measures. The 1-year exposure period was selected to align with exposure periods in previous studies [6, 15]. Example calculations are presented in Online Resource Fig. 1.
Average Daily Dose Measure
The average daily dose measure [6] is calculated by summing the total anticholinergic potency (according to the ACB scale) of all anticholinergics prescribed over the period considered by the ACB scales and dividing the resulting value by the number of days in the period, as follows:
$${\text{Mean}}\,{\text{total}}\,{\text{ACB}}\,{\text{score}}\,{ = }\,\frac{{\sum {\left( {{\text{Drug}}\,{\text{A}}\,\# \,{\text{days}}\,{\text{supplied}}\, \times \,{\text{ACB}}\,{\text{score}}} \right)\,{ + }\,\left( {{\text{Drug}}\,{\text{B}}\,\#\,{\text{days}}\,{\text{supplied}}\,\times {\text{ACB}}\,{\text{score}}} \right)} \,{ + }\,\left( {{\text{Drug}}\,{\text{X }}\ldots } \right)}}{{\#\,{\text{days}}\,{\text{in}}\,{\text{the}}\,{\text{exposure}}\,{\text{period}}}} .$$
Cumulative Dose Measure
To calculate scores based on the cumulative dose measure [5], medication doses are first standardized and then summed to derive an estimate of cumulative exposure, described as the cumulative total standardized daily dose (TSDD) [27, 28]. Steps to calculate TSDD are: (1) calculate total medication dose for each prescription dispensation of a medication considered by an anticholinergic scale like the ACB scale, by multiplying the tablet strength by the number of tablets dispensed; (2) for each prescription dispensation, calculate the standardized daily dose (SDD) by dividing the estimated total medication dose by the minimum-effective dose per day (MED) recommended for use in older adults as per Semla et al.; [29] and (3) for each participant, sum the SDD for all anticholinergic pharmacy dispensations during the exposure period to generate a TSDD. The resulting TSDD is then categorized into “no use” (score of 0); 1–90; 91–365; 366–1095; or greater than 1095, with cut points based on clinical interpretability and the observed exposure distribution.
Cumulative Anticholinergic Burden Measure
The CAB measure was calculated using a novel method based on Gray et al.’s cumulative dose measure, but with the inclusion of medication dosing [5]. Cumulative exposure was thus calculated taking into account both drug-specific properties (i.e., anticholinergic activity) and patient-specific dosing. The World Health Organization (WHO) defined daily dose (DDD), the average daily maintenance dose for a medication’s main indication in adults, was used to standardize dosing across different medications, and the drug-specific ACB score provided strength of anticholinergic activity [30]. Steps to estimate cumulative exposure were: (1) determine the DDD [30, 31] of each medication considered by the ACB scale; (2) calculate the standardized daily dose (SDD) for each anticholinergic dispensing according to the following equation:
$${\text{SDD}}=\frac{{{\text{Number of Daily Units}} \times {\text{Unit Dose}}}}{{\text{DDD}}} .$$
(3) multiply the SDD by the medication’s ACB scale score to yield a drug- and patient-specific measure of standardized daily anticholinergic exposure (SDACE); (4) sum the drug-specific SDACE for all anticholinergic medications for individuals treated with multiple anticholinergic medications on a given day to give a summed standardized daily anticholinergic exposure (SumSDACE); and (5) calculate cumulative exposure by summing SumSDACE for all days during the exposure period. As DDDs are often unavailable in administrative databases, they were extracted from the WHO Collaborating Centre for Drug Statistics Methodology website [32]. An outline of the steps involved in calculating the CAB are provided in Online Resource Figure 3.
Statistical Analysis
Demographic characteristics of the cohort, overall and according to anticholinergic exposure level, were summarized using means and standard deviations (SD) for continuous variables and counts and percentages for categorical variables.
A data-cleaning algorithm was applied to handle clinically implausible values, based on assumptions regarding the most likely data errors responsible, and values adjusted accordingly. Briefly, SDDs above 10 were considered highly unlikely and the following adjustments were made: SDDs ≥ 10 and < 100 were divided by 10, SDDs ≥ 100 and < 1000 were divided by 100, and SDDs ≥ 1000 and < 10,000 were divided by 1000, and so forth.
Measure scores were estimated among those with anticholinergic exposure only. Variability in overall scores for each measure of cumulative anticholinergic exposure was assessed by means (SD) and medians [interquartile range (IQR)]. Note that the absolute values of scores are not directly comparable across measures due to inherent differences in calculation methods, and each measure maps to a different range of values. Spearman’s correlation coefficients were calculated to assess inter-measure correlations of estimated scores.
This article does not contain any new studies with human or animal subjects performed by any of the authors, and, as such, informed consent of individuals was not required. The dataset came from the US-based Truven MarketScan databases from 2012 to 2015.