Potentially Inappropriate Medication Prescribing by Nurse Practitioners and Physicians

Lin-Na Chou MS; Yong-Fang Kuo PhD; Mukaila A. Raji MD; James S. Goodwin MD


J Am Geriatr Soc. 2021;69(7):1916-1924. 

In This Article

Abstract and Introduction


Background: Potentially inappropriate medication (PIM) use is a risk factor for hospitalization and mortality. However, there were few studies focusing on the impact of provider type on PIM use.

Objective: We aimed to estimate the initial and refill PIM prescribing rate for physician visits and nurse practitioner (NP) visits and the impact of provider type on PIM prescribing.

Research Design: We used 100% Texas Medicare data to define physician visits and NP visits in 2016. The rate of visits with a PIM prescription from the same provider was measured, distinguishing between initial and refill prescription to estimate the PIM rate and adjusted odds ratio (OR) by provider type.

Results: There were 24.1 per 1000 visits with a prescription for a PIM: 9.0 per 1000 visits for an initial PIM and 15.1 per 1000 visits for a refill PIM. A visit to an NP was less likely to result in an initial (OR = 0.74, 95% confidence interval [CI] = 0.70–0.79) or refill (OR = 0.54, 95% CI = 0.51–0.57) PIM. The association of lower odds of receiving a prescription for an initial PIM from an NP was substantially stronger among black enrollees than white enrollees (OR = 0.44, 95%CI = 0.30–0.65 for blacks and OR = 0.73, 95%CI = 0.68–0.78 for white enrollees). The association of an NP provider with lower odds of receiving a PIM refill was more pronounced in older patients and in those with more comorbidities.

Conclusions: NPs prescribed fewer initial PIMs and were less likely to refill a PIM after an outpatient visit than physicians. The lower odds of receiving PIMs during an NP visit varied by age, race/ethnicity, rurality, and number of comorbidities.


Adults aged 65 and older are more likely than younger adults to be on five or more medications (polypharmacy), some of which are considered potentially inappropriate or of low value.[1–4] In 2016, 88% of older adults in the United States received one or more prescription medications and 36% had polypharmacy.[4] Polypharmacy is a risk factor for adverse drug events, hospitalization, and mortality.[1,2] Beers criteria[5] were developed in 1991 as a list of potentially inappropriate medications (PIMs) in older patients and has been maintained and updated by the American Geriatrics Society (AGS) since 2013.[6–8] The prevalence of PIM use in the United States ranges from 11.5% to 41.9% among older outpatients[9] and from 4.5% to 33% in the primary care setting.[10]

To increase access to timely primary care, especially in low-income, rural, and other healthcare-disadvantaged populations, nurse practitioners (NPs) have increasingly served as primary care providers (PCPs). Recent changes in the state of NP practice in response to the ongoing shortage of primary care physicians have contributed to the rapid increase of NPs as PCPs in healthcare-disadvantaged areas.[11] Several studies have reported that NPs provide comparable quality and outcomes as physicians for several chronic conditions common in the older adults, such as heart failure, diabetes mellitus, and chronic obstructive pulmonary disease.[12–17] As more NPs serve as PCPs, they assume a greater proportion of prescribing and managing complex medication regimens. A national representative survey data found no difference between primary care physicians and NPs in PIM prescribing, a measure of the quality of prescribing.[18] There are two important limitations of this study. First, it did not control for the important confounding factor of patient characteristics. Second, it did not distinguish between the initial prescription of a PIM and a refill of an existing PIM prescription.

Other studies have focused on the association of patient characteristics and PIM use: female sex, low income, rural residency, taking more medications, and having high healthcare utilization.[19,20] Nevertheless, evidence of the impact of prescriber type on PIM prescribing is limited, and the influence of patient characteristics on PIM prescribing by different types of prescribers has never been evaluated. In addition, previous studies did not differentiate between PIM prescription as an initial prescription and a refill prescription, a difference that would likely influence deprescribing of PIMs. Prescriber attitude and belief are associated with deprescribing an existing medication, especially a PIM that has been used by a patient for years.[21] Also, the medication decision-making and medication selection processes differ between physician and nonphysician prescribers, a reflection of the differences between NPs and physicians in training, background, experience, and practice approach.[22,23]

When a clinician orders a prescription, it is either an initial prescription or a refill authorization for a currently used drug. These are quite different actions in terms of degree of clinical decision-making and responsibility, with initial prescriptions typically occurring after a clinician makes a clinical evaluation and management decision, ideally in the context of shared decision-making. The decision to refill is more likely to be routine, with less consideration. For example, NPs working in a group practice might be asked to refill prescriptions for patients they may not follow, perhaps checking with the primary care physician in questionable cases. Therefore, when considering PIM prescriptions, initial prescriptions should be a more important measure of clinician quality. In this study, we considered initial prescriptions for PIMs separately from refills, with greater emphasis on the initial prescriptions.

We conducted a cross-sectional study using the 2016 Texas Medicare data to compare physician-prescribed PIM and NP-prescribed PIM. Our objectives were twofold: (1) to estimate the initial and refill PIM prescribing rates for physician visits and NP visits and (2) to evaluate the impact of patient characteristics on the association of provider type and PIM rate. We hypothesized that NPs would have less PIM prescribing than physicians and that the odds of receiving a PIM prescription from an NP would differ by patient characteristics.