Physician and patient-related factors associated with inappropriate prescribing to older patients within primary care: a cross-sectional study in Brazil

ABSTRACT BACKGROUND: Physician and patient-related characteristics can influence prescription of medications to older patients within primary healthcare. Use of Brazilian criteria may indicate the real prevalence of prescription of potentially inappropriate medications to this population. OBJECTIVES: To evaluate prescription of potentially inappropriate medications to older patients within primary care and identify patient-related and prescribing physician-related factors. DESIGN AND SETTING: This cross-sectional study was conducted in 22 public primary care facilities in Brazil, among older people (≥ 60 years) who were waiting for medical consultations. METHODS: Interviews were conducted before and after the medical consultations. If the patient received a medical prescription at the consultation, all the drugs prescribed and the physician's medical council registration number were recorded. Prevalence ratios were estimated to ascertain the magnitude of prescription of potentially inappropriate medications, along with patient and physician-related factors associated with such prescription. RESULTS: In total, 417 older patients were included; 45.3% had received ≥ 1 potentially inappropriate medication, and 86.8% out of 53 physicians involved had prescribed ≥ 1 potentially inappropriate medication. The strongest patient-related factor associated with higher prevalence of prescription of potentially inappropriate medications was polypharmacy. Among physician-related factors, the number of patients attended, number of prescriptions and length of medical practice < 10 years were positively associated with prescription of potentially inappropriate medications. CONCLUSIONS: High prevalence of prescription of potentially inappropriate medications was observed. Physician-related characteristics can influence prescription of medications to older people within primary healthcare. This suggests that there is a need for interventions among all physicians, especially younger physicians.


INTRODUCTION
Prescription of potentially inappropriate medications (PIMs) to older adults refers to the prescribing of medications that can potentially increase the risk of adverse health outcomes, given that safer and more effective treatment options are available for a particular indication.
Prescription of PIMs may even refer to a situation in which the risks of therapy outweigh the benefits. [1][2][3][4] Thus, prescribing practices have the biggest impact on occurrences of preventable adverse reactions due to the drug. 3,[5][6][7][8][9] Such occurrences impose significant burdens, like functional decline, falls, delirium and other geriatric syndromes, 5,10,11 and lead to greater numbers of emergency department visits and hospital admissions, and higher mortality rate. 9,12,13 In addition, the high cost of prescribing PIMs to older adults places a great economic burden on healthcare systems. 3,5,8,[14][15][16] Therefore, PIM prescription has become an important worldwide public health issue. 2,3,17 In 2013, 37% of non-hospitalized older people in Canada received ≥ 1 prescription of PIM, equivalent to an estimated cost of US$ 75 per older person or a total cost of US$ 419 million in that year. 15 Similarly, after starting to use at least one PIM, a quarter of older adults continued to use them beyond one year, 18 thereby potentially increasing the risk of adverse events and raising healthcare costs.
Both, clinical and nonclinical factors may influence the prescription of medications. However, it is necessary to recognize factors that may influence clinical decisions, such as physician and patient-related characteristics. These can include the physician's length of medical practice and time in the specialty; and the patient's age, sex, income and comorbidities. 19 In a systematic review of European studies, polypharmacy, advanced age and female sex were the patient-related factors that were most frequently associated with PIM prescription. Other factors that were less frequently associated with PIM prescription included depression, moderate self-rated health quality and poor functional and economic status. 17 In Brazil, a previous study among community-dwelling older adults was conducted to assess the use of PIMs. It was noted that 34.5% of those older people used ≥ 1 PIM and that certain patient factors were associated with the use of inappropriate medication.
These factors included drug use by illiterate older adults, black race, daily use of ≥ 4 medications, use of prescription drugs and acquisition of the drug through the Brazilian public healthcare system as opposed to acquisition in drug stores. 20 However, only a few studies have hitherto evaluated associations of physician-related factors with PIM prescription. 19,[21][22][23] In some of those studies, it was noted that less time spent with patients and certain medical specialties were associated with higher likelihood of PIM prescription. 19,21 Therefore, it is essential to understand the drivers of PIM prescription in order to be able to target specific interventions that might improve prescribing practices for the older adult population.

OBJECTIVE
The aims of this study were to evaluate the frequency of prescription of potentially inappropriate medications to older patients within primary care, in accordance with Brazilian criteria, and to identify the factors associated with such prescriptions, among patients and their prescribing physicians.

Study design
This cross-sectional study constituted the baseline component of a randomized clinical trial named "Development and evaluation of a mobile application for supporting prescription of appropriate medications to the elderly. "

Setting
This study was conducted in all the public primary care facilities in the urban area of Vitória da Conquista (BA), Brazil, which included 15 family health units (FHUs) and 7 primary healthcare units (PHUs). These units belong to the Brazilian public healthcare system (Sistema Único de Saúde, SUS), which provides public healthcare at primary, secondary and tertiary levels to the entire population. Despite the availability of a complementary healthcare system that provides health insurance and private health care for individuals who can afford it, most Brazilian people are assisted only through SUS. The Family Health Strategy is a major program that delivers public primary healthcare to families in areas covered by FHUs. Each FHU has at least one multidisciplinary team that comprises general practitioner physicians, nurses, dentists, nursing technicians and community health workers. The PHUs in the city of Vitória da Conquista provides primary healthcare without the coverage of the Family Health Strategy.

Data collection
Data were collected from all the public primary care facilities in the urban area of the city between September 2016 and December 2017. This was done using a multidimensional questionnaire adapted from the instrument used in the Health, Wellbeing and Aging in Latin America and the Caribbean (SABE) project. 24 This questionnaire was then created in a digital data collection platform (KoBoToolbox, Cambridge, MA, United States).
Participant interviews were conducted both before and after their medical consultations at the 22 primary care facilities mentioned above, and all these medical consultations were timed. After the consultations, if the patient had received a medical prescription, we recorded all drugs that had been prescribed and the physician's registration number in the regional medical council's registry.
If the patient had any health-related interview impediments such as deafness or moderate-to-severe cognitive deficit, all the information was obtained from the person accompanying the patient.

Study population
The patients eligible for inclusion were older people aged ≥ 60 years who were waiting for medical consultations in the facilities. They were invited to participate in the study and were included only after agreeing to this and signing an informed consent form.
We excluded individuals who did not receive a medical prescription and those who had hearing impairments and/or severe cognitive deficits and were unaccompanied by a person with whom the interview could be conducted on the patient's behalf.

Sample description
This study used a non-probabilistic sampling method and the participants were selected through consecutive sampling.
The sample size was estimated at 513 participants, considering a confidence level of 95%, an acceptable difference of 0.05, an assumed proportion of PIM prescription (the event of interest) of 50% and an expected loss rate of 25%.

Variables
The main outcome from the study (dependent variable) was a PIM prescription. The independent variables relating to patients included the following: 1) sociodemographic characteristics (i.e. sex, age, marital status, schooling, income, ethnicity and health insurance); 2) clinical characteristics (i.e. cognitive deficit, sensory deficits [auditory/visual], chronic pain, insomnia, multimorbidity, fall history, functional status and hospitalization); and 3) characteristics relating to medical care (i.e. having a companion attending consultations, consultation length ≤ 10 minutes and prescription of polypharmacy). The independent variables relating to physicians comprised sex, length of time in medical practice, specialty, type of primary care unit, number of patients attended per physician, number of prescriptions per physician and number of PIM prescriptions per physician.

Measurement tools
The prescribed medications were analyzed in terms of the com- 11. Polypharmacy, defined as prescription of ≥ 5 medications. 32 To determine the physician-related variables, the prescribing physicians were firstly identified through their registration numbers in the regional medical council's registry, which was available from the prescription forms that had been issued to the patients.
From these registration numbers, we searched for the following data on the website of the medical council: sex, specialty and year in which the physician issued his or her first prescription. The latter was used to calculate the length of time for which the physician had been in medical practice, in years. The primary care units were classified as FHU or PHU.

Statistical analysis
Descriptive analyses were performed on the variables.
Associations between categorical variables were assessed using the chi-square test, and the prevalence ratio (PR) was measured to estimate the strength of the association. Means of the continuous variables were compared by means of the t test if the statistical assumptions were satisfied, or the Mann-Whitney U test if otherwise.
Multivariable analysis (Poisson regression) was used to adjust for potential confounders. Statistical significance was determined in terms of a 95% confidence interval (CI) or a P-value < 0.05. Data were considered to be missing from the analyses in the following cases: 1) the interviewer marked the option "unknown" in the questionnaire; 2) the physician did not have a registration number in the regional medical council's registry (e.g. if the physician was working in the "More Doctors" program); or 3) certain words and numbers were illegible (e.g. drug name in the prescription and regional medical council registration number). The R statistical software package was used to calculate the prevalence ratio, while all other analyses were undertaken using the SPSS software, version 25 (IBM Corp., Armonk, NY, United States).  Prescriptions provided by 53 physicians were analyzed in this study ( Table 2). Older men accounted for 51% of the physicians; those without a specialty constituted 75.5%; and the median length of medical practice was eight years. A total of 86.8% of the physicians prescribed at least one PIM, with a median of four PIMs prescriptions per physician.

Out
In this study, a total of 1,281 drugs were analyzed, within 417 prescriptions. Forty-two per cent (538/1,281) of the medications prescribed were considered to be PIM-suspect medications. After analysis on the rationale, clinical condition and exceptions, 44.6% (240/538) of them were confirmed as PIMs ( Table 3) According to the univariate analysis in      First-generation antihistamine Dexchlorpheniramine, promethazine 5.4 (13/240) 8 Antispasmodic Scopolamine 3.8 (9/240) 9 Antihypertensive Methyldopa, clonidine, nifedipine 3.8 (9/240) 10 Tricyclic antidepressants Amitriptyline, clomipramine, imipramine 3.8 (9/240) * Dependent variable was PIM; model: (intercept), sex, marital status, insomnia, multimorbidity, polypharmacy and consultation length < 10 minutes; ** Mann-Whitney U test. One minimum monthly wage was equivalent to US$ 283.00. ADL = activity of daily living; CI = confidence interval; IADL = instrumental activity of daily living; PIM = potentially inappropriate medication; PR = prevalence ratio; IQR = interquartile range. The findings from this cross-sectional study warrant future studies to assess interventions that can help improve the appropriateness of drug prescriptions to elderly people. The topics that need to be addressed include education programs targeted to family physicians and the use of decision support systems in primary care settings.

LIMITATIONS
The findings of the present study might be explained by the primary care doctors' unfamiliarity with PIMs and the limited availability of safer therapeutic alternatives for older people through SUS. 47 However, our study was not designed primarily to investigate factors relating to the prescribing physician and, therefore, our sample was not calculated to address these factors.
Moreover, the variables relating to physicians were limited to data held by the Federal Council of Medicine, which therefore left some gaps that future investigations should address.

CONCLUSION
Half of the patients in our current study received ≥ 1 PIM prescription and nine in every ten primary care physicians prescribed ≥ 1 PIM.
Regarding the patient-related factors, polypharmacy was the strongest factor associated with higher prevalence of PIM prescriptions.
Among the physician-related factors, we found that the number of  Last received: November 4, 2020 Accepted: November 18, 2020