Association between Potentially Inappropriate Medication Prescribing and Health-related Quality of Life among U.S. Older Adults

Abstract

Background

Potentially inappropriate medications (PIMs) are associated with worse health outcomes among older adults. Our objective was to examine the association between PIM prescribing and health-related quality-of-life (HRQoL) among older adults in the United States (US) using nationally representative data.

Methods

This was a retrospective study utilizing 2011–2015 Medical Expenditure Panel Survey (MEPS) data. Community dwelling US adults aged 65 years or older were included. A qualified definition operationalized from the 2019 American Geriatrics Society Beers Criteria^® was used to define exposure to PIMs during the study period. The Physical Component Score (PCS) and Mental Components Score (MCS) of the Medical Outcomes Study 12-Item Short Form Health Survey (SF-12) were used to measure HRQoL. Survey-weighted linear regression models were constructed to investigate the association between PIM exposure and participants’ PCS and MCS scores. Analyses were stratified across three age cohorts (65–74, 75–85, and ≥85 years).

Results

Unadjusted analysis showed poorer scores in the PIM exposed group for both PCS and MCS (all P<.001). PIM exposure was associated with poorer PCS scores across all age groups with those aged 65 to 74 years (adjusted regression coefficient = −1.60 [95%CI = −2.27, −.93; P <.001]), those 75 to 84 years (adjusted regression coefficient: −1.49 [95%CI = −2.45, −.53; P =.003]), and those 85 years and older (adjusted regression coefficient = −1.65 [95%CI = −3.03, −0.27; P = .02]). PIM exposure was also associated with poorer MCS scores in participants aged 65 to 74 years (adjusted regression coefficient = −0.69 [95%CI = −1.16, −0.22; P =.004]) and 85 years and older (adjusted regression coefficient = −2.01 [95%CI = −3.25, −0.78; P =.002]).

Conclusions

Our results suggest that patient’s exposure to PIMs is associated with poorer HRQoL. Further work is needed to assess whether interventions to deprescribe PIMs may help to improve patient’s HRQoL.

INTRODUCTION

Potentially inappropriate medication (PIM) prescribing and polypharmacy continue to rise as an individual ages, along with the number of comorbidities.¹ Older adults above the age of 65 represent roughly 17% of the US population; however, they account for the most medication use.² PIMs are defined as medications where the potential for harm outweighs its benefits and they account for 20% of ADEs older adults experience, and simply taking a PIM doubles the likelihood of an avoidable ADE.³ Additionally, these ADE events result in increased healthcare utilization and costs for older adults, as up to 10% of all hospital admissions among older adults are ADE-related. ^4–6

PIMs also influence health-related quality of life (HRQoL). HRQoL is defined as an individual or group’s perceived physical and mental health over time.⁷ Measuring HRQoL has become an important public health surveillance tool as it is an indicator of unmet needs and intervention outcomes. Lower HRQoL is also associated with increased mortality and morbidity.^8,9 Several studies have investigated the relationship between PIMs and their effects on HRQoL of patients. Several studies have found that PIMs increase healthcare utilization and reduce HRQoL. ^10–12 However, other studies have found no relationship between PIMs and HRQoL using the same or similar methods. ^13,14 A recent meta-analysis examining PIM prescribing and its effects found that although there was an association between PIMs and HRQoL, the pooled data had high heterogeneity reducing the overall level of evidence.¹⁵ Much of the current evidence comes from relatively small studies conducted outside of the United States and therefore may not be generalizable to the current population and contemporary prescribing patterns.^10,11,13

The purpose of this study was to assess the relationship of PIM prescribing with HRQoL in a U.S. nationally representative sample using the 2019 American Geriatrics Society (AGS) Beers criteria^®.¹⁶

METHODS

Data Source and Population

We conducted a cross-sectional analysis using data collected from 2011 to 2015 MEPS. MEPS captured information from sample household that participated in the prior year’s National Health Interview Survey to provide a nationally representative sample of the U.S. civilian noninstitutionalized population.¹⁷ We used the MEPS Full-Year Consolidated files, Prescribed Medicines files, and the Medical Conditions files from the 2011–2015 study years. We included only those who were aged 65 or older during the first round of interview. This study was conducted using publicly available data, therefore no ethical approval was necessary.

Exposure to PIM

Data regarding prescription drugs were collected through household questionnaire for participants and confirmed through electronic records from pharmacies outlining prescription type, dosage, and payment. The primary exposure was defined as any fill for a PIM based on the 2019 AGS Beers Criteria^® which consisted of 33 of the 35 medications or classes of medications that were considered generally inappropriate for older adults.¹⁶ A qualified definition of exposure was adapted using the approach described by Davidoff et al. that has been described previously. ^6,18 Details on the definition of PIM exposure are described in more detail in Supplementary Table S1.

Outcomes

The main outcome of our study was patient’s HRQoL assessed using the 12-Item Short Form Health Survey (SF-12) from the Full-Year Consolidated files. The SF-12 is a 12-item self-administered questionnaire in which participants are asked to respond about their physical and mental health status.¹⁹ Responses to these questions are weighted to compute the Physical Component Summary (PCS) and Mental Component Summary (MCS) scores. The questionnaire contains eight domains including physical functioning, role-physical, bodily pain, general health, vitality, social functioning, role-emotional, and mental health. Questions in each domain contribute to both scores. Both the PCS and MCS were standardized using a scale of 0 (worst) to 100 (best) with a mean of 50 and standard deviation of 10 to represent the health status of each participant. Participants with a score of greater or less than 50 indicated that their health status was better or worse than the general U.S. average, respectively. The SF-12 has been validated in multiple populations including MEPS participants.^20,21

Statistical Analysis

Participant’s baseline demographics and comorbidities were compared between PIM users and nonusers using the chi-squared test. Stratification of age groups facilitated the comparison of SF-12 scores which accounts for the age-related differences in functional status. The HRQoL scores of study participants, represented by PCS and MCS scores, were compared between PIM users and nonusers using independent t-tests. Linear regression models were constructed to investigate the association between PIM exposure and participants’ PCS and MCS scores with 95% CIs. To further investigate the dose-response relationship, we performed additional analysis by replacing the binary variable with categorical variables for 0 PIM, 1 PIM, and 2 or more PIMs. All models were adjusted for covariates associated with PIM prescribing. Information on covariates can be found in Supplement S1. Survey weighted procedures were used in the HRQoL analysis to produce national estimates and account for the complex survey design. All statistical analyses were 2-tailed with a level of significance set at P <0.05 utilizing SAS version 9.4 (SAS Institute, Cary, NC).

RESULTS

This study included 218,383,123 participants aged 65 and older from 2011 to 2015, of whom 75,135,061 (34.4%) were prescribed at least one PIM. Among PIM users, 24,854,304 (33.1%) participants were prescribed 2 or more PIMs. The characteristics of the population are described further in Table 1.

Table 1.

Study Population Characteristics between PIM and Non-PIM Prescribing Groups within the Medical Expenditure Panel Survey, 2011 – 2015^1,2

Characteristic	Total n=218,383,123	SE	Non-PIM Exposure n=143,248,062	SE	PIM Exposure n=75,135,061	SE	P
Age							.027
65-74	56.9	0.8	57.8	0.9	55.3	1.1
75-84	30.9	0.7	29.9	0.8	32.7	0.9
85+	12.2	0.6	12.3	0.6	12.0	0.8
Sex							.005
Female	55.9	0.5	54.8	0.9	58.1	0.9
Race/Ethnicity							<.001
White (Non-Hispanic)	77.8	1.0	76.8	1.2	79.9	1.0
Black (Non-Hispanic)	8.7	0.5	9.1	0.6	7.8	0.6
Hispanic	7.7	0.5	7.9	0.6	7.3	0.6
Other	5.9	0.7	6.3	0.9	5.1	0.6
Marital Status							.04
Married	55.4	0.9	56.3	1.0	53.6	1.3
Education							<.001
High school or less	49.9	1.0	49.1	1.1	51.5	1.3
Some college	23.2	0.6	22.5	0.7	24.7	0.9
College or postgraduate	26.8	0.9	28.5	0.9	23.8	1.2
Income (% poverty line)							<.001
High income	38.1	1.0	39.7	1.0	35.0	1.2
Middle income	28.4	0.6	27.9	0.6	29.4	0.8
Low income	17.6	0.5	17.3	0.6	18.1	0.7
Near poor	6.3	0.3	5.6	0.3	7.5	0.4
Poor/negative	9.7	0.4	9.4	0.4	10.1	0.5
Insurance							<.001
Medicare only	36.0	0.8	36.4	0.9	35.2	1.0
Medicare and private	52.1	0.9	52.2	1.0	52.0	1.1
Medicare and public	10.8	0.5	10.0	0.5	12.3	0.8
No Medicare/uninsured	1.1	0.1	1.4	0.2	0.6	0.1
ADL limitations							<.001
Yes	6.6	0.3	5.7	0.3	8.1	0.5
IADL limitations							<.001
Yes	10.9	0.4	9.4	0.4	13.9	0.8
Census region							<.001
Northeast	19.2	1.0	20.6	1.1	16.5	1.0
Midwest	22.3	0.9	21.7	1.0	23.4	1.2
South	37.0	1.3	35.6	1.3	39.7	1.6
West	21.4	0.9	22.0	1.1	20.3	0.9
Comorbidities
Coronary heart disease	19.5	0.5	17.1	0.6	23.9	0.9	<.001
Angina	7.5	0.3	6.0	0.3	10.2	0.6	<.001
Myocardial infarction	12.3	0.4	10.8	0.4	15.0	0.7	<.001
Heart failure	2.7	0.2	2.2	0.2	3.7	0.4	<.001
Chronic renal failure	0.5	0.1	0.5	0.1	0.7	0.1	.097
Cancer	30.9	0.7	29.2	0.8	34.1	1.0	<.001
Arthritis	59.1	0.6	53.3	0.7	70.1	0.9	<.001
Hypertension	68.8	0.6	65.7	0.7	74.6	0.9	<.001
Dyslipidemia	62.5	0.6	59.6	0.8	68.1	1.0	<.001
Asthma	8.7	0.3	7.7	0.3	10.6	0.6	<.001
Stroke	12.3	0.4	10.7	0.4	15.4	0.6	<.001
Emphysema	6.4	0.3	5.5	0.3	8.3	0.5	<.001
Chronic bronchitis	4.4	0.2	3.8	0.2	5.7	0.4	<.001
Diabetes	22.4	0.5	18.3	0.6	30.1	0.9	<.001
Anxiety	11.4	0.4	4.9	0.4	23.7	0.9	<.001
Mood disorder	12.4	0.4	7.6	0.4	21.6	0.8	<.001
Dementia	5.0	0.3	4.2	0.3	6.4	0.5	<.001
Number of Medications
0-4	43.3	0.7	56.3	0.8	18.5	0.7	<.001
5 or more	56.7	0.7	43.7	0.8	81.5	0.7
Number of PIMs							–
1	23.0	0.6	–	–	66.9	1.0
2	8.6	0.3	–	–	25.0	0.8
3 or more	2.8	0.2	–	–	8.1	0.5

Abbrev: SE, standard error; PIM, potentially inappropriate medication; ADL, activities of daily living; IADL, instrumental activities of daily living.

¹All percentage are weighted to provide estimates for approximately 218 million older adults in the United States between 2011 and 2015.

²Differences were compared between PIM users and non-users with the χ² test

The PIM user group had lower HRQoL scores than the nonuser group in terms for both PCS and MCS (Figure 1). After adjusting for potential confounders, PIM exposure was associated with poorer PCS scores across all age groups with those aged 65 to 74 years (adjusted regression coefficient = −1.60 [95% CI = −2.27, −.93; P <.001]), those 75 to 84 years (adjusted regression coefficient: −1.49 [95% CI = −2.45, −.53; P =.003]), and those 85 years and older (adjusted regression coefficient = −1.65 [95% CI = −3.03, −0.27; P = .02]) (Supplementary Table S2). PIM exposure was also associated with poorer MCS scores in participants aged 65 to 74 years (adjusted regression coefficient = −0.60 [95% CI = −1.16, −0.22; P =.004]) and 85 years and older (adjusted regression coefficient = −2.01 [95% CI = −3.25, −0.78; P =.002]).

Figure 1.

Mean SF-12 health-related quality of life scores based on PIM exposure. PIM, Potentially inappropriate medication; SF-12, 12-Item Short Form Health Survey. (A) Patients aged 65–74 years. (B) Patients aged 75–84 years. (C) Patients aged ≥85 years.

Comparison of HRQoL scores revealed significant differences in HRQoL scores among groups with different number of prescribed PIMs (Table 2). Exposure to 1 or 2 PIMs was significantly associated with lower PCS scores among participants across all age groups with a larger effect seen in those with ≥2 PIM, with the exception of those over 85 years old (adjusted regression coefficient = −1.33 [95% CI = −2.81, 0.14; P =.076]). Significantly lower MCS scores were noted among those aged 65 to 74 years prescribed ≥2 PIMs (adjusted regression coefficient = −1.07 [95% CI = −1.93, −0.23; P = .013]), and those ≥85 years old prescribed 1 PIM (adjusted regression coefficient = −2.26 [95% CI = −3.67, −0.85; P = .002]).

Table 2.

Associations between Number of PIM and Health-Related Quality of Life Scores

	Physical Component Summary (PCS) of SF-12 1
Age	Unadjusted Coefficient1 (95% CI)	P	Adjusted Coefficient1,2 (95% CI)	P
65-74  0 (reference)  1  ≥2	– -4.56 (-5.38, -3.73) -7.88 (-9.07, -6.68)	– <.001 <.001	– -1.20 (-1.92, -0.48) -2.56 (-3.56, -1.56)	– .001 <.001
75–84  0 (reference)  1  ≥2	– -3.02 (-4.03, -2.01) -5.69 (-7.37, -4.01)	– <.001 <.001	– -1.03 (-2.02, -0.03) -2.60 (-4.03, -1.18)	– .043 .005
≥85  0 (reference)  1  ≥2	– -2.79 (-4.33, -1.25) -6.07 (-8.32, -3.82)	– <.001 <.001	– -1.33 (-2.81, 0.14) -2.60 (-4.95, -0.25)	– .076 .031
	Mental Component Summary (MCS) of SF-12 1
Age	Unadjusted Coefficient 1 (95% CI)	P	Adjusted Coefficient1,2 (95% CI)	P
65-74  0 (reference)  1  ≥2	– -2.31 (-2.91, -1.70) -4.33 (-5.45, -3.20)	– <.001 <.001	– -0.53 (-1.07, 0.02) -1.07 (-1.93, -0.23)	– .057 .013
75–84  0 (reference)  1  ≥2	– -1.32 (-2.34, -0.31) -3.02 (-4.27, -1.77)	– .011 <.001	– 0.16 (-0.96, 0.99) -0.03 (-1.19, 1.25)	– .973 .961
≥85  0 (reference)  1  ≥2	– -3.91 (-5.41, -2.41) -4.70 (-6.59, -2.82)	– <.001 <.001	– -2.26 (-3.67, -0.85) -1.27 (-3.20, 0.66)	– .002 .196

Abbreviations: SF-12, 12-item short form health survey; CI, confidence interval; PIM, potentially inappropriate medication.

¹Survey weights applied to give a national estimate

²Adjusted models controlled for age, sex, race, marital status, education, income, insurance coverage, ADL limitations, IADL limitations, geographic region, coronary heart disease, angina, myocardial infarction, heart failure, chronic renal failure, cancer, arthritis, hypertension, dyslipidemia, asthma, stroke, emphysema, chronic bronchitis, diabetes, and dementia, anxiety, mood disorders, and polypharmacy.

DISCUSSION

Our findings suggest that PIMs, identified by the 2019 AGS Beers Criteria^®, are negatively associated with HRQoL among U. S. older adults. The use of at least one PIM was significantly associated with lower HRQoL in participants among all age groups studied. The impact on PCS was even larger when participants were prescribed ≥ 2 PIMs. MCS seemed to be most significantly impacted in those aged ≥85 years old, suggesting that this population may benefit most from interventions to reduce PIM prescribing.

Our results support the findings of several previous studies that have established a negative relationship between PIMs and HRQoL.^10,11 Cahir et al. found that PIMs, identified by the Screening Tool of Older Persons’ Prescriptions (STOPP) criteria, were associated with significant adverse health outcomes. Participants with multiple PIMs reported lower HRQoL per the European Quality of Life 5 Dimension (EQ-5D) instrument, similar to the results in our study.¹¹ Wallace et al. found that participants prescribed ≥ 2 PIMs had statistically significant reductions in EQ-5D scores as well.¹⁰ Both studies demonstrated that any PIM use was associated with a reduction in HRQoL with use of ≥ 2 PIMs leading to greater reductions in HRQoL scores. Our study found similar results using SF-12 scores. However, unlike these two studies, our study used nationally representative data making it more generalizable to older adults in the US. Our study builds upon the study by Fu et al, which utilized 1996 MEPS data and found a significant reduction in self-perceived general health status due to PIM use.¹² The authors adjusted their models for multiple confounding variables and the negative relationship persisted. Our study found similar results between PIM use and HRQoL utilizing more recent MEPS data from 2011–2015.

Contrasting the results of our study, Moriarty et al. found no difference between quality of life among participants prescribed PIMs and no PIMs.¹³ This difference in results may be attributed to the use of the Control, Autonomy, Self-Realization, and Pleasure revised scale (CASP-R12). This scale is a revised version of the CASP-19 and measures the overall quality of life of the participant using a social model, not a medical one.²² The survey itself has very few questions pertaining to the participant’s health, reducing its applicability to HRQoL.²³ Additionally, our results counter those of Franic and Jiang.¹⁴ These authors also utilized MEPS data from 2000–2001 and applied the 2003 Beers criteria and found no association between PIM use and adverse HRQoL. Despite analyzing 5 separate outcome measures for HRQoL, no association was found in the regression analysis. This difference in results may be attributed to several factors including differences in study population and differences in medications included in the PIM exposure.

Our study demonstrates that PIM use among older U. S. adults is associated with a reduced HRQoL. Multiple interventions, such as deprescribing, comprehensive medication reviews, nursing education, and geriatric assessments, have been studied to reduce PIM use in older adults and improve their HRQoL.²⁴ Romskaug et al. demonstrated that assessment and medication review by a geriatrician and follow up with a family physician could improve HRQoL among home-dwelling patients with polypharmacy.²⁵ Pitkala et al. implemented nursing education at assisted living facilities as the intervention to reduce inappropriate medication use and examined its effect on HRQoL, and found that improved education not only significantly reduced the number of harmful medications used by the nurses, but also slowed the decline in HRQoL for the patients.²⁶ While these studies provide some evidence that reducing PIM exposure may improve HRQoL, most studies have been limited in their sample size and generalizability. Therefore, large cohort studies and randomized control trials involving PIMs deprescribing or similar interventions need to be conducted to examine their effects on HRQoL among older U. S. adults. Future studies are warranted to confirm these results in a more contemporary US population using updated versions of the AGS Beers Criteria and STOPP criteria.^27,28

Limitations

This study has several limitations. While our methods for defining PIMs were robust, our methods did not allow for a temporal relationship to be assessed between the exposure and the outcome. Although we used the qualifications outlined by the criteria, some participants deemed as PIM users may have been prescribed the medication appropriately based on a diagnosis not recorded in the MEPS database. Additionally, we used truncated ICD-9-CM or CCS codes to examine patient conditions which may have also limited our identification of inappropriate medication use in some participants. While we accounted for many demographic factors and comorbidities, the possibility of residual confounding remains. The data used for this study are not the most current, which may limit the generalizability of the findings to the current US population. We maintain that the study results remain relevant regarding PIM exposure and its relationship with HRQoL. However, further studies are warranted to validate these findings using more up to date data. Additionally, we categorized multiple PIMs within the same class as a single PIM. This not only underestimates the prevalence of PIMs, but also underestimates the effects multiple PIMs from within the same class may have on older adults. Lastly, MEPS does not collect any data pertaining to medication purchased over the counter, so prevalence of some medications may be underestimated.

CONCLUSION

Our study demonstrates that the use of PIMs in older U. S. adults is associated with a poorer HRQoL. Having more than one PIM was generally associated with a poorer HRQoL. Further work is necessary to assess whether interventions, such as deprescribing PIMs, in this population can improve the patient’s quality of life.

Supplementary Material

Supinfo

Supplement S1 Provides a detailed description of covariates included in analysis.

Supplementary Table S1 Includes the qualifications used to define medications as potentially inappropriate used as the exposure in this analysis.

Supplementary Table S2 Displays the results of our regression analysis analyzing exposure to any PIM and HRQoL scores.

Key Points.

• Prescribing of potentially inappropriate medications was associated with lower health-related quality of life scores among community dwelling older adults.

• The impact appears to be great among patients prescribed more than one class of potentially inappropriate medications.

Why does this matter?

This paper helps to establish a relationship between potentially inappropriate medications and health-related quality of life among community dwelling older adults in the United States that can be used to benchmark this outcome and inform future interventional studies on deprescribing these medications.