If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Aging Research Center, Karolinska Institutet, Stockholm, SwedenEpiChron Research Group on Chronic Diseases, Aragon Health Sciences Institute, IIS Aragon, Miguel Servet University Hospital, Zaragoza, Spain
The burden of medications near the end of life has recently come under scrutiny, because several studies suggested that people with life-limiting illness receive potentially futile treatments.
Methods
We identified 511,843 older adults (>65 years) who died in Sweden between 2007 and 2013 and reconstructed their drug prescription history for each of the last 12 months of life through the Swedish Prescribed Drug Register. Decedents' characteristics at time of death were assessed through record linkage with the National Patient Register, the Social Services Register, and the Swedish Education Register.
Results
Over the course of the final year before death, the proportion of individuals exposed to ≥10 different drugs rose from 30.3% to 47.2% (P <.001 for trend). Although older adults who died from cancer had the largest increase in the number of drugs (mean difference, 3.37; 95% confidence interval, 3.35 to 3.40), living in an institution was independently associated with a slower escalation (β = −0.90, 95% confidence interval, −0.92 to −0.87). During the final month before death, analgesics (60.8%), anti-throm-botic agents (53.8%), diuretics (53.1%), psycholeptics (51.2%), and β-blocking agents (41.1%) were the 5 most commonly used drug classes. Angiotensin-converting enzyme inhibitors and statins were used by, respectively, 21.4% and 15.8% of all individuals during their final month of life.
Conclusion
Polypharmacy increases throughout the last year of life of older adults, fueled not only by symptomatic medications but also by long-term preventive treatments of questionable benefit. Clinical guidelines are needed to support physicians in their decision to continue or discontinue medications near the end of life.
The burden of medications increases near the end of life (47.2% of older adults receive ≥10 prescription drugs during their last month of life).
•
Polypharmacy is fueled not only by symptomatic medications but also by long-term preventive treatments of questionable benefit.
•
Guidelines are needed to support physicians in their decision to continue or discontinue drug treatments near the end of life.
Under the combined effect of increased longevity, chronic multimorbidity, and single-disease clinical guidelines, the concomitant use of multiple medications has become commonplace among older adults.
The relationship between number of drugs and potential drug-drug interactions in the elderly: a study of over 600,000 elderly patients from the Swedish Prescribed Drug Register.
When considering older people near the end of life, poly-pharmacy poses 2 problems. First, as death approaches, age-related physiologic changes are amplified by the changing metabolism, the decline of renal and hepatic functions, and the loss of body mass. As a result, pharmacokinetics and pharmacodynamics are altered, making older adults with life-limiting illness particularly vulnerable to the harmful side effects of medications.
Second, the accumulation of prescriptions in the context of limited life expectancy raises questions about the intended or expected benefit of the treatments.
The burden of medications near the end of life has recently come under scrutiny, because several studies have shown that people with specific life-limiting diseases are prescribed medications whose benefit is unlikely to be achieved within their remaining lifespan.
Future research and clinical guidelines need to be informed by findings that are generalizable beyond a specific illness or care setting.
This study aimed to measure the change in the prevalence of polypharmacy and to identify the most commonly used medications over the course of the last year of life of older people, using data with national coverage in Sweden.
Methods
Study Design and Population
We conducted a nationwide, follow-back cohort study of all older adults who died at age >65 years in Sweden between January 1, 2007 and December 31, 2013. Individuals were excluded from the study population if they had no reported cause of death or had no prescription data available during the final 3 months before the date of death (Supplementary Figure 1, available online). Death certificate data were obtained from the Swedish National Board of Health and Welfare and were linked at the individual level to several other registries with national coverage: the Swedish Prescribed Drug Register, Social Services Register, National Patient Register, and Swedish Education Register. Data were anonymized, and the Regional Ethical Review Board in Stockholm approved the study (no. 2013/1941-31/3 and 2015/1319-32).
Assessment of Polypharmacy and Drug Exposure
Polypharmacy was considered as the primary outcome. Although there is no consensual definition, studies conducted in the general geriatric population typically use a threshold of ≥4 or ≥5 medications to characterize polypharmacy and ≥9 or ≥10 to describe “excessive polypharmacy.”
we opted for the latter, more conservative cut-off. Hence polypharmacy was hereafter defined as the monthly exposure to 10 or more prescription drugs, that is, distinct substances according to the fifth level of the Anatomical Therapeutic Chemical (ATC) classification system.
Prescription drug data were derived from the Swedish Prescribed Drug Register to evaluate the total number of medications during each of the final 12 months before death. In Sweden, drug prescriptions cover a maximum period of 90 days. Drug exposure was estimated according to 1) the date of dispensing, 2) the total amount dispensed to the patient, and 3) the prescribed daily dose, as described in Supplementary Figure 2 (available online).
In addition, we calculated the prevalence of the 20 most common individual drug classes for the 12th, 6th, and final months before death. As recommended by the World Health Organization, drugs were classified by ATC code.
Descriptive Variables
Sex and age at time of death were both extracted from death certificates. International Classification of Diseases, 10th revision diagnosis codes for all contributing causes of death were categorized into 4 distinct “illness trajectories” indicative of the potential timeframe of care needs near the end of life: cancer, organ failure, prolonged dwindling, and sudden death.
(Supplementary Table 1, available online). When multiple causes of death indicated more than 1 illness trajectory, we applied a predefined hierarchy (ie, from cancer to prolonged dwindling to organ failure to sudden death).
to provide a comprehensive account of the burden of disease near the end of life (Supplementary Table 2, available online). In brief, a total of 60 chronic diseases were identified through the National Cause of Death Register (all causes of death), the National Patient Register (inpatient and specialized outpatient diagnoses reported during the last 24 months before death), and the Swedish Prescribed Drug Register (use of specific medications in the last 24 months before death). Living arrangement of the individuals during the final year of life was categorized as “community-dwelling” or “institutionalized,” using data from the Social Services Register. Level of education was used as a measure of socioeconomic position and operationalized as “primary education,” “secondary education,” and “tertiary education.”
Statistical Analysis
Descriptive analysis was performed by calculating the proportion and means of the decedents' characteristics according to their living arrangement. Statistical differences were tested with Pearson's χ2 for categorical variables and median test for continuous variables, with P <.001. The proportion of individuals receiving ≥10 drugs was reported together with the mean number of prescription drugs throughout the final year of life. Results comparing community-dwelling and institutionalized individuals were standardized for both sex and age by using the total population as reference.
We investigated trends in polypharmacy and in the number of prescribed drugs between the 12th month and the final month before death. Unadjusted logistic regression models were computed to calculate the odds of being exposed to polypharmacy during the final month before death compared with 12 months before, using clustered robust standard error to account for intra-individual correlation. We also identified factors independently associated with change in the number of prescribed drugs over time by means of generalized estimating equation models with identity link functions at the individual level and unstructured correlation. The number of distinct medications was considered as a continuous dependent variable, whereas sex, age, living arrangement, number of chronic diseases, illness trajectory, and level of education were entered as independent variables. Results are presented as adjusted β coefficients with their 95% confidence intervals (CIs). In sensitivity analyses the same models were computed by removing analgesics (ATC code N02) from the total number of prescription drugs, because this class of medications is used for specific purposes near the end of life. Finally, changes in the prevalence of individual drug classes were assessed with variation rates (with 95% CI) from the 12th to the final month before death. Bivariate logistic regressions were used to assess P for trend (2-sided α = .01). All analyses were performed with SAS JMP 12.1 (SAS Institute, Cary, NC) and Stata 14.1 (StataCorp, College Station, TX) software.
Results
Study Population
Of all 545,212 older adults who died in Sweden between 2007 and 2013, 511,843 (93.9%) met our eligibility criteria. Table 1 shows the decedents' main characteristics. Compared with community dwellers, institutionalized individuals were more often women, died at older age, and were more likely to follow a trajectory of prolonged dwindling. Overall, 51% of individuals had 5 or more diagnosed chronic conditions at time of death. The most common chronic illnesses were ischemic heart diseases (42%), hypertension (40%), congestive heart failure (38%), cancer (31%), atrial fibrillation (27%), and cerebrovascular diseases (26%).
Table 1Characteristics of Older Adults Who Died Between 2007 and 2013 in Sweden, by Living Arrangement
Illness trajectories were constructed using all the diagnoses mentioned on the death certificates (ie, both underlying and contributing causes of death). When the different causes of death indicated more than 1 illness trajectory, we applied a previously described hierarchy (ie, from cancer to prolonged dwindling to organ failure to sudden death).25
n (%)
<.001
Cancer
121,569 (34.6)
23,309 (14.5)
144,878 (28.3)
Organ failure
143,840 (41.0)
54,758 (34.0)
198,598 (38.8)
Prolonged dwindling
56,514 (16.1)
72,243 (44.9)
128,757 (25.2)
Sudden death
29,054 (8.3)
10,556 (6.6)
39,610 (7.7)
No. of chronic diseases
Mean (SD)
5.0 (2.7)
4.7 (2.7)
4.9 (2.7)
<.001
n (%)
<.001
0
3192 (0.9)
2352 (1.5)
5544 (1.1)
1
21,116 (6.0)
13,020 (8.1)
34,136 (6.7)
2
39,361 (11.2)
21,137 (13.1)
60,498 (11.8)
3
49,957 (14.2)
24,215 (15.1)
74,172 (14.5)
4
52,361 (14.9)
23,904 (14.9)
76,265 (14.9)
≥5
184,990 (52.7)
76,238 (47.4)
261,228 (51.0)
SD = standard deviation.
∗ Pearson's χ2 test, except for the mean age and mean number of chronic conditions (median test).
† Missing values for level of education: 24,319 (4.7% of the total cohort).
‡ Illness trajectories were constructed using all the diagnoses mentioned on the death certificates (ie, both underlying and contributing causes of death). When the different causes of death indicated more than 1 illness trajectory, we applied a previously described hierarchy (ie, from cancer to prolonged dwindling to organ failure to sudden death).
The prevalence of polypharmacy increased significantly throughout the last year of life (Figure). Between the 12th and the final month before death, the proportion of individuals exposed to ≥10 different drugs rose from 30.3% to 47.2% (P <.001 for trend), and the mean (standard deviation) number of prescription drugs increased from 7.6 (4.4) to 9.6 (4.7). Older adults living in institutions were found to receive a greater number of medications than those living in the community (P <.001 for each of the final 12 months of life). These trends remained after excluding analgesics from the total count of prescription drugs (Supplementary Figure 3, available online). However, we found significant differences across age groups and illness trajectories (Supplementary Figure 4, available online).
Figure(A) Polypharmacy and (B) number of prescription drugs over the course of the last 12 months of life of older people in Sweden, by living arrangement. Proportions and means are standardized for both sex and age, using the total population as reference. All trends statistically significant (P <.01).
Factors Associated with Change in the Number of Prescription Drugs
The magnitude of the increase in polypharmacy over the course of the last year of life varied significantly according to the decedents' characteristics (Table 2). Hence, the likelihood of being exposed to polypharmacy during the final month of life compared with 12 months before death was greater for individuals who died from cancer (odds ratio 3.34; 95% CI, 3.29-3.39) than for those who died from dementia or other neurodegenerative disorders (odds ratio 1.74; 95% CI, 1.71-1.77). As shown in Table 3, female sex, older age at time of death, and institutionalization were independently associated with a slower increase in the average number of medications over time. Older adults who died from cancer had the greatest increase in the number of prescription drugs (mean difference 3.37; 95% CI, 3.35-3.40). These associations were only partly attenuated when analgesics were excluded from the total count of prescribed drugs (Supplementary Tables 3 and 4, available online).
Table 2Change in the Likelihood of Being Exposed to Polypharmacy During the Last Year of Life of Older Adults in Sweden, 2007-2013
Odds ratio calculated by means of logistic regression models including polypharmacy (≥10 prescription drugs) as a binary dependent variable and time (last month before death vs. 12th month before death) as sole independent variable. Clustered standard errors at the individual level were used to account for correlation of observations within groups.
Illness trajectories were constructed using all the diagnoses mentioned on the death certificates (ie, both underlying and contributing causes of death). When the different causes of death indicated more than 1 illness trajectory, we applied a previously described hierarchy (ie, from cancer to prolonged dwindling to organ failure to sudden death).25
Cancer
144,878
24.3
51.8
112.8 (110.5-115.9)
3.34 (3.29-3.39)
Organ failure
198,598
37.0
50.5
36.3 (35.3-37.3)
1.73 (1.71-1.76)
Prolonged dwindling
128,757
28.4
40.9
43.8 (42.3-45.4)
1.74 (1.71-1.77)
Sudden death
39,610
23.9
34.3
43.2 (40.0-46.4)
1.66 (1.61-1.71)
No. of chronic diseases
0
5544
8.3
12.8
54.7 (38.4-72.9)
1.63 (1.44-1.84)
1
34,136
8.9
18.6
109.8 (101.5-118.5)
2.35 (2.24-2.46)
2
60,498
13.3
26.2
97.1 (92.3-141.9)
2.32 (2.25-2.39)
3
74,172
18.0
33.6
86.9 (83.5-92.4)
2.31 (2.25-2.37)
4
76,265
22.8
41.3
81.1 (78.3-84.4)
2.38 (2.33-2.44)
≥5
261,228
43.1
62.1
44.0 (43.2-44.7)
2.16 (2.14-2.18)
CI = confidence interval.
∗ Variation rates indicate the relative change in prevalence of polypharmacy between the 12th month and the final month before death.
† Odds ratio calculated by means of logistic regression models including polypharmacy (≥10 prescription drugs) as a binary dependent variable and time (last month before death vs. 12th month before death) as sole independent variable. Clustered standard errors at the individual level were used to account for correlation of observations within groups.
‡ Missing values for level of education: 24,319 (4.7% of the total cohort).
§ Illness trajectories were constructed using all the diagnoses mentioned on the death certificates (ie, both underlying and contributing causes of death). When the different causes of death indicated more than 1 illness trajectory, we applied a previously described hierarchy (ie, from cancer to prolonged dwindling to organ failure to sudden death).
β-Coefficients computed by the mean of generalized estimating equation models with identity link functions at the individual level and unstructured correlations. The total number of distinct prescription drugs (5-digit Anatomical Therapeutic Chemical codes) was entered as continuous dependent variable with Gaussian distribution, while including all presented individual characteristics as independent variables. β-Coefficients for the interaction with time can be interpreted as the rate of change between the 12th month and the final month before death, compared with the reference category.
Illness trajectories were constructed using all the diagnoses mentioned on the death certificates (ie, both underlying and contributing causes of death). When the different causes of death indicated more than 1 illness trajectory, we applied a previously described hierarchy (ie, from cancer to prolonged dwindling to organ failure to sudden death).25
Cancer
144 878
6.7 (4.4)
10.0 (4.9)
3.37 (3.35-3.40)
Ref
Organ failure
198 598
8.4 (4.5)
10.0 (4.8)
1.59 (1.57-1.61)
−1.79 (−1.82 to −1.77)
Prolonged dwindling
128 757
7.5 (3.9)
8.9 (4.1)
1.36 (1.34-1.38)
−2.01 (−2.04 to −1.98)
Sudden death
39 610
6.8 (4.2)
8.1 (4.4)
1.29 (1.26-1.31)
−2.09 (−2.14 to −2.05)
No. of chronic diseases
0
5544
4.6 (3.2)
5.6 (3.3)
0.91 (0.85-0.97)
Ref
1
34 136
4.8 (3.4)
6.4 (3.6)
1.60 (1.57-1.64)
0.74 (0.61-0.86)
2
60 498
5.4 (3.6)
7.3 (3.8)
1.86 (1.83-1.89)
0.97 (0.86-1.09)
3
74 172
6.1 (3.7)
8.1 (4.0)
1.97 (1.94-1.99)
1.08 (0.96-1.20)
4
76 265
6.8 (3.8)
8.9 (4.1)
2.08 (2.06-2.11)
1.19 (1.07-1.30)
≥5
261 228
9.1 (4.4)
11.2 (4.7)
2.12 (2.10-2.14)
1.22 (1.10-1.33)
CI = confidence interval; SD = standard deviation.
∗ β-Coefficients computed by the mean of generalized estimating equation models with identity link functions at the individual level and unstructured correlations. The total number of distinct prescription drugs (5-digit Anatomical Therapeutic Chemical codes) was entered as continuous dependent variable with Gaussian distribution, while including all presented individual characteristics as independent variables. β-Coefficients for the interaction with time can be interpreted as the rate of change between the 12th month and the final month before death, compared with the reference category.
† Missing values for level of education: 24,319 (4.7% of the total cohort).
‡ Illness trajectories were constructed using all the diagnoses mentioned on the death certificates (ie, both underlying and contributing causes of death). When the different causes of death indicated more than 1 illness trajectory, we applied a previously described hierarchy (ie, from cancer to prolonged dwindling to organ failure to sudden death).
Antithrombotic agents, diuretics, analgesics, psycholeptics, and β-blocking agents were found to be the 5 most common prescription drug classes during the last year of life (Table 4). Agents acting on the renin–angiotensin system, antianemic preparations, antidepressants, and drugs for acid-related disorders were also highly prevalent, with more than 30% of decedents exposed to these drugs during the final month before death. The pattern of prescription drug use changed over the course of the last 12 months of life, with a notable increase in the exposure to opioids (+120.7%), antimicrobials (+74.3%), anxiolytics (+59.5%), drugs for constipation (+57.8%), and antipsychotics (+47.3%). We found only a modest decrease in the use of preventive drugs. Hence, during their last month of life, a significant share of older adults used β-blockers (41.1%), angiotensin-converting enzyme inhibitors (21.4%), vasodilators (17.4%), lipid-lowering agents (16.3%), calcium channel blockers (15.4%), or potassium-sparing agents (12.1%). The prevalence of preventive medication was found to be higher among younger individuals (Supplementary Table 5, available online).
Table 4Prevalence of the 20 Most Commonly Used Prescription Drug Classes During the Last Year of Life of Older People in Sweden, 2007-2013
Drug classes are ranked by descending order, using the second level of the ATC (therapeutic subgroups, eg, N02 Analgesics). For some classes, details are also provided for the pharmacological subgroups (eg, N02A Opioids).
ATC Code
Prevalence, % (N = 511,843)
Variation Rate Between 12th Month and Last month Before Death, % (95% CI)
Variation rates calculated as the relative increase or decrease in prevalence of use between the 12th month and the final month before death, in with their 95 CI.
12th Month
6th Month
Last Month
1. Antithrombotic agents
B01
52.5
53.5
53.8
2.6 (2.2-2.9)
Vitamin K antagonists
B01AA
7.9
7.7
6.7
−15.4 (−16.6 to −14.3)
Heparin group
B01AB
1.6
2.5
5.2
221.6 (213.8-229.6)
Platelet aggregation inhibitors
B01AC
44.3
44.9
44.9
1.4 (0.9-1.8)
2. Diuretics
C03
47.1
49.2
53.1
12.7 (12.3-13.2)
Low-ceiling diuretics
C03A-C03B
6.0
5.7
5.2
−12.5 (−13.9 to −11.1)
High-ceiling diuretics
C03C
37.2
40.1
45.6
22.7 (22.1-23.3)
Potassium-sparing agents
C03D
9.6
10.3
12.1
26.0 (24.6-27.4)
3. Analgesics
N02
40.2
45.9
60.8
51.2 (50.6-51.8)
Opioids
N02A
17.5
21.4
38.5
120.7 (119.1-122.2)
Other analgesics
N02B
35.0
39.6
49.2
40.6 (40.0-41.3)
4. Psycholeptics
N05
39.5
42.8
51.2
29.6 (29.0-31.1)
Antipsychotics
N05A
7.8
8.9
11.5
47.3 (45.6-49.1)
Anxiolytics
N05B
16.7
18.8
26.6
59.5 (58.3-61.7)
Hypnotics and sedatives
N05C
28.1
29.9
34.5
23.1 (22.4-23.8)
5. β-Blocking agents
C07
39.4
40.3
41.1
4.1 (3.6-4.6)
6. Drugs acting on the renin–angiotensin system
C09
31.8
31.9
30.6
−3.9 (−4.5 to −3.4)
ACE inhibitors
C09A-C09B
21.6
21.9
21.4
−1.0 (−1.7 to −0.2)
Angiotensin II receptor antagonists
C09C-C09D
10.9
10.7
9.8
−9.9 (−11.0 to −8.9)
7. Anti-anemic preparations
B03
30.6
32.9
34.6
12.9 (12.3-13.5)
Iron preparations
B03A
7.8
8.9
10.4
33.6 (32.0-35.3)
Vitamin B12 and folic acid
B03B
25.9
27.5
28.2
9.0 (8.3-9.7)
8. Psychoanaleptics
N06
27.4
29.5
32.6
18.9 (18.2-19.6)
Antidepressants
N06A
24.5
26.6
30.1
22.7 (21.9-23.5)
Anti-dementia drugs
N06D
5.4
5.4
5.0
−8.2 (−9.7 to −6.6)
9. Drugs for constipation
A06
25.3
30.1
39.9
57.8 (56.9-58.7)
10. Drugs for acid related disorders
A02
23.8
27.4
35.1
47.3 (46.4-48.2)
11. Cardiac therapy
C01
22.2
23.0
24.3
9.3 (8.4-19.4)
Cardiac glycosides
C01A
7.9
8.0
8.3
5.1 (3.7-6.5)
Vasodilators used in cardiac diseases
C01D
15.5
16.2
17.4
12.5 (11.6-13.5)
12. Emollients and protectives
D02
22.1
25.3
28.7
29.9 (29.0-35.8)
13. Lipid modifying agents
C10
18.7
18.1
16.3
−12.9 (−13.6 to −12.1)
Statins
C10AA
18.2
17.6
15.8
−13.3 (−14.1 to −12.6)
14. Calcium channel blockers
C08
17.8
17.1
15.4
−13.0 (−13.8 to −12.3)
15. Mineral supplements
A12
17.6
18.7
20.5
16.8 (15.9-17.8)
Calcium
A12A
12.6
13.0
12.9
2.8 (1.8-3.8)
Potassium
A12B
5.6
6.3
8.1
45.3 (43.2-47.4)
16. Drugs used in diabetes
A10
15.1
15.3
15.4
2.2 (1.3-3.1)
Insulin and analogues
A10A
8.8
9.4
10.3
16.5 (15.1-17.9)
Oral blood glucose lowering drugs
A10B
8.6
8.2
7.4
−14.0 (−15.1 to −12.8)
17. Ophthalmologicals
S01
14.7
15.2
15.3
3.8 (2.9-4.8)
18. Drugs for obstructive airway diseases
R03
12.5
13.3
14.9
19.5 (18.3-25.7)
19. Antibacterials for systemic use
J01
11.5
13.1
20.0
74.3 (72.6-75.9)
20. Thyroid therapy
H03
10.5
10.8
10.9
4.4 (3.3-5.6)
ACE = angiotensin-converting enzyme; ATC = Anatomical Therapeutic Chemical classification system; CI = confidence interval.
∗ Drug classes are ranked by descending order, using the second level of the ATC (therapeutic subgroups, eg, N02 Analgesics). For some classes, details are also provided for the pharmacological subgroups (eg, N02A Opioids).
† Variation rates calculated as the relative increase or decrease in prevalence of use between the 12th month and the final month before death, in with their 95 CI.
Our study has two main findings. First, the proportion of older adults exposed to ≥10 different prescription drugs increases over the course of the last year before death, and approximately half of individuals experience polypharmacy during their last month of life. Second, polypharmacy is fueled not only by an increased use of medications directed toward symptom management but also by the frequent continuation of long-term preventive treatments and disease-targeted drugs.
These results compare to earlier studies. McNeil et al
recently reported that patients with a life-limiting disease took on average 10.7 medications at time of death, with 69% of patients using 9 or more different drugs (excluding statins). In a retrospective cohort study of 100 patients who died from advanced cancer, researchers found a median of 11 prescribed medications 9 days before death.
Despite a considerable heterogeneity in study designs and populations, other studies investigating drug use near the end of life have reported comparable results.
Polypharmacy near the end of life should, however, not be thought of as a homogeneous phenomenon. Data presented in our article suggest that polypharmacy does not stem from a unique drug class but is in fact propelled by 5 different types of prescriptions, each addressing specific goals of care. Medications to alleviate the burden of symptoms (eg, analgesics, loop-diuretics, anxiolytics) compose the first of these 5 categories. This is expected, because the need for comfort care and symptom management increases sharply as death approaches.
The second category includes drug regimens used in the long-term prevention of chronic diseases that pose no immediate danger. Hence we found that statins were rarely discontinued and were still prescribed to 16% of older adults during their last month of life. Although this figure is consistent with previous reports,
it raises serious concern: the clinical benefit of treatments drugs aiming at preventing cardiovascular diseases during the final month of life is at the very least questionable.
Medications used to control the evolution of potentially life-threatening or disabling comorbidities form a third distinct group of prescriptions (eg, oral antidiabetics, platelet aggregation inhibitors, thyroid therapy, antidementia drugs, and cardiac stimulants). The fourth group is drugs prescribed in an attempt to cure or slow the progression of the main life-limiting illness (eg, chemotherapy, immunosuppressants). Finally, drugs administered to counteract the current or anticipated adverse effects of other medications constitute a fifth group of prescriptions. The widespread use of potassium supplementation, proton-pump inhibitors, and laxatives close to death may reflect a “prescribing cascade” (ie, prescriptions prompted by the onset of symptoms induced by other medications rather than by the disease itself).
recommended that 2 key questions be taken into account. First, is the patient's life expectancy longer than the time needed for the medication to achieve its benefit? Second, are the objectives of the prescribed medication in keeping with the goals of care that the physician and the patient discussed and agreed upon? This prescribing model relies on the idea that drug treatments should be adapted to mirror the course of the disease as the remaining life expectancy diminishes. In other words, physicians should consider discontinuing drugs that may be effective and otherwise appropriate but whose potential harms outweigh the benefits that patients can reasonably expect before death occurs. The process of deprescribing (ie, withdrawing medications with the aim of improving health outcomes
In a recent randomized, controlled trial including patients with ≤12 months of estimated life expectancy, researchers found no significant survival difference according to statin continuation/discontinuation and showed that quality of life was significantly improved in the group of patients that discontinued statins.
Whether this conclusion is applicable to other drug classes remains uncertain, because data regarding the safety of discontinuing preventative medications in the context of end-of-life care remain scarce.
One should also balance the potential lack of benefit of a certain drug with the possible collateral effects of its discontinuation. For instance, although there is considerable uncertainty about the efficacy and safety of antidementia agents in older adults reaching the end of life,
and by the natural tendency of healthcare professionals and patients to believe that medical interventions are more effective than they actually are (a phenomenon described as the “therapeutic illusion”
). The frequent continuation of potentially futile treatments may also be the consequence both of a lack of communication and of insufficient shared decision making between patients and prescribers, out of fear that discussing issues surrounding the end of life would be perceived as a failure or that it would amount to “giving up.”
To overcome this predicament, the process of deprescribing requires timely patient–family–physician dialogue about the risk/benefit ratio of medications, and close monitoring of symptoms during the following weeks. As the situation worsens, the goals of care should be frequently reassessed to ensure that the treatment target remains concordant with the patient's preferences. It is also essential that patients and their relatives receive clear information about their options in terms of palliative care, to counter the feeling of abandonment that they may experience when disease-directed treatments are withdrawn.
Our study should be interpreted with caution, considering the following limitations. First, the Swedish Prescribed Drug Register only records prescription drugs dispensed at pharmacies. Over-the-counter medications and medications administered in hospitals or from drug store rooms in nursing homes are not reported in the register, leading to a potential underestimation of the actual drug exposure. However, approximately 86% of defined daily doses dispensed to the Swedish population annually are prescription drugs delivered through pharmacies and are therefore covered by the Swedish Prescribed Drug Register. Second, we could only estimate drug exposure with the assumption that patients use the dispensed medications at the prescribed rate; we could therefore not account for nonadherence to treatment. Additionally, we did not investigate whether drugs that were continued until the final month of life were deintensified in terms of dosage, which could correspond to a shift toward a palliative approach. Notwithstanding these methodologic restrictions, it is the first study investigating the burden of medications near the end of life in an entire population (ie, irrespective of the underlying disease or care setting).
Conclusions
Polypharmacy increases sharply throughout the last year of life of older adults. To reduce the burden of medications of questionable benefit in older adults with life-limiting illness, robust evidence about the benefit and safety of deprescribing is needed. Following the example set by Kutner et al
with the statin discontinuation trial, future clinical trials should be conducted to evaluate the effects of withdrawing preventive medications in people with advanced illness. These findings should then be embedded into clinical guidelines, in the same manner that new evidence regarding the benefit and safety of initiating drug therapy is incorporated into current practice guidelines. However, because end-of-life situations are shaped by different disease trajectories, symptoms, and personal preferences, the goals of care vary considerably from one person to another. Future clinical practice guidelines should thus foster personalized decision making rather than promote the systematic discontinuation of medications according to a one-size-fits-all set of criteria.
Acknowledgments
We thank Régis Aubry, Jonas Wastesson, and Emerald Heiland for their insightful comments on the manuscript.
Supplementary Figure 2Design of drug exposure calculation. Tick marks represent the date of drug dispensing (“start date”). Horizontal lines represent the estimated duration of drug exposure. Arrows represent the end of drug exposure (“end date”).
Supplementary Figure 3Change in the number of nonanalgesic prescription drugs over the course of the last 12 months of life, by living arrangement. Standardized for both sex and age (reference: total population). ATC = Anatomical Therapeutic Chemical classification system.
Supplementary Figure 4Polypharmacy (≥10 Drugs) over the course of the last 12 months of life of older people in Sweden, by illness trajectory and age at time of death.
Supplementary Table 2Details of Diagnosis Codes and Drugs Used to Detect Chronic Conditions in the Last 2 Years of Life of Older People in Sweden
Adapted with the authors' permission from Calderón-Larrañaga A, Vetrano DL, Onder G, et al. Assessing and measuring chronic multimorbidity in the older population: a proposal for its operationalization. J Gerontol A Biol Sci Med Sci. 2016 December 21.
Chronic Disease
ICD-10 Codes
ATC Codes
Source(s) of data
National cause of death register (all contributing causes of death) National patient register (all inpatient and specialized outpatient diagnoses)
Supplementary Table 3Change in the Likelihood of Being Exposed to ≥10 Prescription Drugs (Excluding Analgesics) During the Last Year of Life of Older Adults in Sweden, 2007-2013
Variation rates indicate the relative change in prevalence of polypharmacy (≥10 prescription drugs excluding analgesics) between the 12th month and the final month before death.
Odds ratio computed by the mean of logistic regression models including polypharmacy (≥10 prescription drugs excluding analgesics) as a binary dependent variable and time (last month before death vs 12th month before death) as sole independent variable. Clustered standard errors at the individual level were used to account for correlation of observations within groups.
Illness trajectories were constructed using all the diagnoses mentioned on the death certificates (ie, both underlying and contributing causes of death).
Cancer
144,878
19.1
36.5
91.7 (89.3-94.2)
2.45 (2.40-2.49)
Organ failure
198,598
31.3
42.1
34.8 (33.6-35.9)
1.60 (1.58-1.62)
Prolonged dwindling
128,757
21.6
28.9
33.8 (32.0-35.6)
1.48 (1.45-1.50)
Sudden death
39,610
18.8
26.0
38.3 (34.7-42.3)
1.52 (1.47-1.57)
No. of chronic diseases
0
5544
4.5
6.6
44.8 (23.9-69.3)
1.48 (1.25-1.75)
1
34,136
5.4
9.5
77.2 (67.7-87.3)
1.85 (1.75-1.97)
2
60,498
8.9
15.2
71.7 (66.3-77.2)
1.85 (1.78-1.91)
3
74,172
12.6
21.6
71.2 (67.2-75.2)
1.91 (1.86-1.96)
4
76,265
17.1
28.9
69.3 (66.1-72.6)
1.98 (1.93-2.02)
≥5
261,228
36.4
51.0
40.0 (39.1-46.9)
1.82 (1.80-1.84)
∗ Variation rates indicate the relative change in prevalence of polypharmacy (≥10 prescription drugs excluding analgesics) between the 12th month and the final month before death.
† Odds ratio computed by the mean of logistic regression models including polypharmacy (≥10 prescription drugs excluding analgesics) as a binary dependent variable and time (last month before death vs 12th month before death) as sole independent variable. Clustered standard errors at the individual level were used to account for correlation of observations within groups.
‡ Missing values for level of education: 24,319 (4.7% of the total cohort).
§ Illness trajectories were constructed using all the diagnoses mentioned on the death certificates (ie, both underlying and contributing causes of death).
Supplementary Table 4Factors Associated with Change in the Number of Prescription Drugs (Excluding Analgesics) During the Last Year of Life of Older Adults in Sweden, 2007-2013
β-Coefficients computed by the mean of generalized estimating equation models with identity link functions at the individual level and unstructured correlations. The total number of distinct prescription drugs excluding analgesics was entered as continuous dependent variable with Gaussian distribution, while including all presented individual characteristics as independent variables. β-Coefficients for the interaction with time can be interpreted as the rate of change between the 12th month and the final month before death, compared with the reference category.
Illness trajectories were constructed using all the diagnoses mentioned on the death certificates (ie, both underlying and contributing causes of death).
Cancer
144,878
6.1 (4.0)
8.3 (4.3)
2.24 (2.22-2.26)
Ref
Organ failure
198,598
7.8 (4.2)
9.0 (4.4)
1.24 (1.22-1.25)
−1.01 (−1.03 to −0.99)
Prolonged dwindling
128,757
6.8 (3.6)
7.7 (3.8)
0.84 (0.83-0.86)
−1.39 (−1.42 to −1.37)
Sudden death
39,610
6.2 (3.9)
7.2 (4.0)
1.00 (0.97-1.02)
−1.25 (−1.29 to −1.21)
No. of chronic diseases
0
5544
4.1 (2.9)
4.6 (2.9)
0.57 (0.52-0.63)
Ref
1
34,136
4.2 (3.0)
5.2 (3.1)
1.03 (1.00-1.05)
0.48 (0.37-0.59)
2
60,498
4.9 (3.2)
6.1 (3.3)
1.23 (1.21-1.26)
0.67 (0.57-0.78)
3
74,172
5.6 (3.4)
6.9 (3.5)
1.35 (1.32-1.37)
0.78 (0.68-0.89)
4
76,265
6.3 (3.5)
7.7 (3.6)
1.46 (1.44-1.48)
0.98 (0.78-0.99)
≥5
261,228
8.4 (4.1)
9.9 (4.2)
1.51 (1.50-1.53)
0.93 (0.83-1.04)
CI = confidence interval; SD = standard deviation.
∗ β-Coefficients computed by the mean of generalized estimating equation models with identity link functions at the individual level and unstructured correlations. The total number of distinct prescription drugs excluding analgesics was entered as continuous dependent variable with Gaussian distribution, while including all presented individual characteristics as independent variables. β-Coefficients for the interaction with time can be interpreted as the rate of change between the 12th month and the final month before death, compared with the reference category.
† Missing values for level of education: 24,319 (4.7% of the total cohort).
‡ Illness trajectories were constructed using all the diagnoses mentioned on the death certificates (ie, both underlying and contributing causes of death).
Supplementary Table 5Prevalence of the 20 Most Commonly Used Prescription Drug Classes During the Last Month of Life of Older People in Sweden, by Age (2007-2013)
Drug classes are ranked by descending order, using the second level of the ATC (therapeutic subgroups, eg, N02 Analgesics). For some classes, details are also provided for the pharmacologic subgroups (eg, N02A Opioids).
Prevalence, %
66-74 y (N = 75,749)
75-84 y (N = 163,355)
85-94 y (N = 227,391)
≥95 y (N = 45,348)
1. Antithrombotic agents (B01)
44.2
54.9
57.5
48.0
Vitamin K antagonists (B01AA)
7.0
9.1
5.9
1.5
Heparin group (B01AB)
9.9
5.9
3.6
2.5
Platelet aggregation inhibitors (B01AC)
30.9
43.5
50.5
45.3
2. Diuretics (C03)
39.5
49.7
58.6
60.3
Low-ceiling diuretics (C03A-C03B)
4.9
5.8
5.2
3.9
High-ceiling diuretics (C03C)
31.5
41.8
51.4
54.1
Potassium-sparing agents (C03D)
11.9
12.2
12.3
11.5
3. Analgesics (N02)
55.7
58.4
63.0
66.9
Opioids (N02A)
42.7
38.8
37.2
37.0
Other analgesics (N02B)
38.9
46.0
53.2
57.7
4. Psycholeptics (N05)
48.4
49.7
52.6
54.7
Antipsychotics (N05A)
9.4
11.2
12.2
12.0
Anxiolytics (N05B)
26.5
25.6
26.8
29.2
Hypnotics and sedatives (N05C)
34.0
33.8
35.1
35.5
5. β-Blocking agents (C07)
37.7
43.5
42.5
30.4
6. Drugs acting on the renin-angiotensin system (C09)
∗ Drug classes are ranked by descending order, using the second level of the ATC (therapeutic subgroups, eg, N02 Analgesics). For some classes, details are also provided for the pharmacologic subgroups (eg, N02A Opioids).
The relationship between number of drugs and potential drug-drug interactions in the elderly: a study of over 600,000 elderly patients from the Swedish Prescribed Drug Register.
Funding: This work was supported by a grant from the Swedish Research Council for Health, Working Life and Welfare (No 2014-4699 ). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Conflict of Interest: None.
Authorship: LM conceived and designed the study, performed the statistical analysis, interpreted the data, drafted, and critically revised the manuscript. DLV, DR, and AC-L and interpreted the data and critically revised the manuscript. JF acquired, analyzed, and interpreted the data and critically revised the manuscript. KJ obtained funding, provided supervision, interpreted the data and, critically revised the manuscript. LM is the guarantors of the study and data integrity. All authors gave approval for the final version of the manuscript and agree to be accountable for all aspects of the work.
Ethical Approval: This study was approved by the Ethical Review Board in Stockholm, Sweden.
Availability of Data and Materials: Clinical data and individual data from the Swedish Prescribed Drugs Register data cannot be made publicly available. Interested researchers can access the aggregated data from the Swedish Prescribed Drugs Register (http://www.socialstyrelsen.se/statistik/statistikdatabas/lakemedel).
Polypharmacy is a major public health problem in the United States and abroad. Although there is no standard definition of polypharmacy, I see and recognize it every day in the outpatient and inpatient environment. A typical example that I saw yesterday was an 84-year-old man admitted to our internal medicine service for worsening heart failure. He had a list of 14 different medicines that he claimed to be ingesting each day. This list even included 2 different β-blockers! In addition, he was taking significant quantities of an over-the-counter nonsteroidal anti-inflammatory agent, which had caused deterioration of his renal function and was a major factor in his worsening heart failure.
30253-X&id=gr1.jpg){kind=link}
30253-X&id=fx1.jpg){kind=link}
30253-X&id=fx2.jpg){kind=link}
30253-X&id=fx3.jpg){kind=link}
30253-X&id=fx4.jpg){kind=link}