How Chronic Is Polypharmacy in Old Age?

A Longitudinal Nationwide Cohort Study

Jonas W. Wastesson, PhD; Lucas Morin, MSc; Marie-Laure Laroche, PhD; Kristina Johnell, PhD


J Am Geriatr Soc. 2019;67(3):455-462. 

In This Article


Study Population

We used register data with nationwide coverage in Sweden to create a longitudinal cohort of older adults (aged 65 years or older) who were exposed to five or more drugs in October 2010. Study participants were followed up prospectively until December 2013 (ie, for up to 37 months). The Swedish Prescribed Drug Register was linked to the National Patient Register, the National Cause of Death Register, and the Social Services Register, as described elsewhere.[24] We excluded individuals who died during the first 12 months of follow-up, as people at the end of life might have specific clinical needs.[25] The selection of the study population is presented in Supplementary materials Figure S1.

Figure S1.

Study population flowchart

Outcome Measurement: Polypharmacy

Data regarding prescription drug use were extracted from the Swedish Prescribed Drugs Register, which collects information about all prescription drugs delivered in pharmacies in Sweden.[26] Exposure periods were constructed for each dispensed drug based on: (i) the date of drug dispensing, (ii) the number of dispensed defined daily doses, and (iii) the prescribed daily dose, as reported by the prescriber.[27,28] We then calculated the number of different drugs used in each 30-day window (ie, distinct substances according to the fifth level of the Anatomical Therapeutic Chemical [ATC] classification system). As illustrated in Figure S2, individuals were considered as exposed to polypharmacy during a given month when the number of drugs was five or more.

Figure S2.

Calculation of monthly drug exposure in the Swedish Prescribed Drug Register
Note. Tick marks represent the date of drug dispensing ("start date"), horizontal lines represent the estimated duration of drug exposure, and arrows represent the end of drug exposure ("end date")

To distinguish "chronic" from "transient" polypharmacy exposure, we used the different approaches illustrated in Figure 1. Health problems are usually defined as "chronic" when they persist over time without any measurable interruptions (eg, diabetes, heart failure). To reflect this, we calculated the duration of polypharmacy as the number of consecutive months spent with five or more different drugs. We considered the first episode, starting at baseline and stopping when the patient was no longer exposed to polypharmacy for at least 2 months. In other words, interruptions in polypharmacy exposure were discarded if they lasted 1 month or less. This "grace period" was used to reduce the influence of irregular drug refill patterns. The chronicity of polypharmacy was calculated as the proportion of individuals who remained exposed for 6 and 12 months or longer.

Figure 1.

Fictitious example of two persons followed up from baseline until the end of the study period (ie, for a follow-up time of 37 months in total). Each square represents 1 month. The washout period of 6 months before baseline is used to distinguish persons who were already exposed to polypharmacy before baseline (person A) from those who started a new polypharmacy episode at baseline (person B). Each episode of polypharmacy starts at the first month of exposure and ends when the person remains unexposed for at least 2 consecutive months (grace period). In this example, both persons are considered as having a first episode of polypharmacy that persisted for 7 months, followed by two other episodes of polypharmacy. The fraction of time with polypharmacy is calculated as the number of months with polypharmacy, including grace periods, divided by the total number of months of available follow-up. In this example, the fraction of time with polypharmacy is equal to 33/37 (89.2%). Thus, considering a cutoff value of 80% or greater, these persons are defined as chronic polypharmacy users.

Other health problems are characterized by short interruptions during their progression, but they can still be considered as chronic if people are experiencing them more often than not (eg, chronic pain, psoriasis). The underlying assumption is that some conditions occur so frequently that their impact on people's everyday life is constant, although their onset appears as a series of discrete events. To mirror this second scenario, we calculated the fraction of time with polypharmacy by dividing the number of months with polypharmacy (numerator) by the total number of months of available follow-up (denominator). The numerator did include grace periods. We then defined chronic polypharmacy users as older adults who had a fraction of time with polypharmacy of 80% or greater (eg, at least 30 months of 37 for those surviving the complete follow-up). This is similar to how drug adherence is calculated using the medication possession ratio.[29]

Other Covariates

Living arrangement at baseline was defined as "community dwelling" or "living in institution," using data from the Social Services Register. Multimorbidity was assessed using a validated assessment tool,[5] which captures 60 distinct chronic diseases using data from the national patient register during the 3 years prior to baseline, as well as data about specific medications dispensed during the same period. This variable was defined as the number of chronic conditions, with five or more conditions as the maximum value. Multidose dispensing (in Swedish, ApoDos) refers to drugs administered through portion-packed plastic pouches. It is especially common among older adults living in nursing homes in Sweden.[30]

Statistical Analysis

We calculated the duration of polypharmacy for each individual and identified those who remained exposed for 6 and 12 or more consecutive months. To account for left censoring, we stratified the population according to their exposure to polypharmacy during the 6-month period before baseline. Since we excluded older adults who died during the first year of follow-up, outcome measurement was not affected by right censoring (ie, survival). However, the persistence of polypharmacy throughout the entire follow-up was analyzed with Kaplan-Meier survival functions accounting for mortality. We then measured the fraction of time with polypharmacy as the number of months spent with polypharmacy divided by the total number of months of available follow-up. The proportion of older adults who had a fraction of time with polypharmacy of 80% or greater was reported with percentages. Since this indicator is proportional to the contributing time of each individual, it is not affected by mortality selection. We analyzed factors associated with a high fraction of time with polypharmacy using multivariate logistic regression models adjusted for age, sex, living arrangement, number of chronic conditions, dispensing regimen, and number of drugs at baseline. All estimates from the logistic regression are calculated as predicted probabilities and presented as percentages (with 95% confidence intervals) using the margins command in Stata version 14.1 (StataCorp, College Station, TX). Predicted probabilities can be compared across models and can be interpreted as adjusted proportions conditional on the covariates.[31] Post hoc, we stratified the analysis by dispensing regimen to investigate the combined effect of living arrangement and dispensing regimen. In sensitivity analyses, the fraction of time with polypharmacy was categorized using a lower cutoff value (50% instead of 80%), which has previously been used as a definition of chronic polypharmacy.[32]

Ethical Approval

Data were anonymized, and the Regional Ethical Review Board in Stockholm approved the study (2013/1941-31/3 and 2015/1319-32).