Mapping Drug Prescription, Polypharmacy, and Pharmaceutical Spending in Older Adults in Iran: A Multilevel Analysis Based on Claims Data.

Background: To date, comprehensive data on drug utilization in Iranian people are lacking. The purpose of this study was to graphically describe drug prescription, polypharmacy, and pharmaceutical spending in > 3 million Iranian elderly people.
Methods: In this multilevel cross-sectional study, using administrative claims data provided by the Iran Health Insurance Organization (IHIO), we assessed drug claims and drug costs in 2018 in >3 million individuals living in Iran and who have been insured with health insurance (Bimeh Salamat). In particular, we analyzed the prevalence of polypharmacy and pharmaceutical spending use according to the annual Report of Iranian Health Insurance Organization. Multilevel ordinal logistic and multilevel beta regression models were used to analyze the data. Significance level was set as P ≤ .05 and CI at 95%.
Results: Nationally, the mean number of drug prescriptions per patient was 1.46 (SD, 0.81). The mean number of prescribed drugs per patient was 4.32 (SD, = 3.04). The drug cost for each elderly patient was $6.86 (interquartile range (IQR), 12.26), with $4.96 and $1.76 for the insurance and the insured shares, respectively. For elderly women, the odds of polypharmacy (excessive and nonexcessive vs no polypharmacy) were 1.164 (95% CI, 1.142 to 1.186) times that of elderly men. In addition, in the spring season, the odds of polypharmacy were 1.274 (95% CI, 1.241 to 1.309) times that of the winter. Similarly, polypharmacy was strongly higher among patients who had noncommunicable diseases (OR, 2.174; 95% CI, 2.069 to 2.275 (P < 0.001)).
Conclusion: The high prevalence of hyper prescription in Iran elderly people may indicate a need for interventions aiming at deprescribing drugs with an unfavorable benefit-risk profile. The best practice guidelines should be developed for improved medical practice in the prescription of medications for such a vulnerable population.

What is “already known” in this topic

Concomitant use of multiple prescription drugs (polypharmacy) and inappropriate prescribing are important clinical challenges, especially among older patients who often have multiple chronic conditions.

What this article adds

The high prevalence of hyper prescription in Iran elderly people may indicate a need for interventions aiming at deprescribing drugs with an unfavourable benefit-risk profile. Best practice guidelines should be developed for improved medical practice in the prescription of medications for such a vulnerable population.

Introduction

Drug therapy is the most commonly used method of any disease treatment in general practice (1). Concomitant use of multiple prescription drugs (polypharmacy) and inappropriate prescribing are important clinical challenges, especially among older patients who often have multiple chronic conditions. Polypharmacy is heterogeneously defined in studies as the use of multiple medications by the patient, although most studies agree on a threshold of at least 5 medications (2, 3). In Iran, one-quarter of community-dwelling patients aged >65 years self-report taking 5 or more drugs (4), although higher rates have been reported based on claims data (5- 7). Demographic transition and population aging are one of the major health challenges in the 21st century. According to official censuses, Iran, like many European countries, is aging. In 1996, about 6.6% of the total population of Iran were people over 60 years old, which reached 7.7% in 2006, and it is projected to reach about 10% of the total population by 2021 (8). This aging process is a multidimensional human, social, economic, cultural, and health issue (9, 10). The elderly are inclined to many diseases because of physiological changes and aging. The incidence of various diseases in the elderly causes these people to take more drugs than other periods of life and, as a result, suffer more drug side effects. Decreased cardiac output, decreasing liver and renal function, hemodynamic changes, along with drug interactions, and the use of over-the-counter medications aggravate the clinical condition of the elderly (11).

Drugs are also expensive and account for 25% of all health care expenditures (12), thus, their rational and efficient use is essential. The appropriate use of medicines can achieve better and safer health care for patients and communities (13, 14). The World Bank estimates that 20% to 50% of health care costs in developing countries are spent on medicines and medical equipment (13). In addition, some studies show that polypharmacy was substantially increased with the increasing insurance share in spending costs (15, 16).

Several studies have reported that polypharmacy is associated with certain risk factors such as increasing age (17- 20), female gender (18- 20), body mass index (19, 21), and the number of comorbidities (17, 20, 21). Other studies have also shown that polypharmacy is associated with certain noncommunicable diseases (NCDs), such as diabetes mellitus, hypertension, cardiovascular diseases, asthma, and dyslipidemia (20- 23).

Studies from different countries have reported varying rates polypharmacy in the elderly, for instance, ranging from 18 % in Brazil, 44% in Sweden, and 86% in South Korea (24- 26). In the United States, polypharmacy has tripled over 2 decades to reach 39% (27). One primary reason for this variation may be that there is no clear universal definition for this phenomenon. Studies have used different definitions.

A related phenomenon to the high prevalence rates in polypharmacy is that the elderly population is increasing globally (28). The elderly population in Iran is also increasing due to public health initiatives and improved health care services across the country (29).

Associated with aging is the increasing prevalence of polypharmacy, which necessitates the need for pharmaceutical spending (30, 31). Therefore, investigating the prevalence of polypharmacy and its association with age and sex is crucial to implement measures that promote the rational use of medication. Thus, the objective of this study was to determine and map the drug prescriptions, polypharmacy (>5 drugs), and pharmaceutical spending among elderly patients ( > 65 years) and its association with demographic characteristics in more than 3 million elderly Iranian patients using health insurance claim data.

Methods

Study Design and Data Source

We conducted a retrospective descriptive analysis of health insurance claims data covering the years 2018-2019. Data for analysis were extracted using the Electronic Medical Record (EMR) database from a large-scale anonymized health insurance claims database, obtained from National Center for Health Insurance Research (NCHIR) for senior citizens aged ≥65 years, which manages the health insurance program (Bime Salamat) for Iranian residents. Data were obtained from April 2018 to March 2019. As of March 2019 (Esfand 1397 in the Solar Hijri calendar), the insurance program was covering 3,039,629 beneficiaries aged ≥65 years from 429 cities (nested in 31 provinces) in Iran.

The data retrieved from the EMR included patients’ demographic and clinical characteristics, such as age, gender, number of drug prescription, number of prescribed drugs, season of drug prescription, total pharmaceutical spending, insurance share, and insured share. Also, the most frequent NCDs in this population were selected. These conditions included each of the following: diabetes, asthma, dyslipidaemia, hypertension, gastrointestinal reflux disease, cardiovascular diseases (ischemic heart disease, heart failure, arrhythmia, and stroke), arthritis (osteoarthritis and rheumatoid arthritis), and mental health conditions (depression, anxiety and dementia).

Definitions

Drug prescription: The number of drug prescription to each patient was quantified by counting the prescriptions that were administered monthly and regularly prescribed for any diseases in an elderly patient setting during the 12-month study period.

Furthermore, the elderly with stable disease conditions might visit outpatient clinics only once every 2 to 3 months. As a result, if only 1 month of data were analyzed, we would overlook elderlies who had no scheduled outpatient visits in that particular month. Consequently, we analyzed 12 months of data to ensure the inclusion of all 3-month prescriptions for elderly patients without monthly clinic visits. Therefore, we categorized prescriptions in three levels: 1 prescription, 2 to 3 prescriptions, and equal to and more than 4 prescriptions.

Polypharmacy: At present, there is no international consensus on the definition of polypharmacy. Based on previous studies (3), we defined polypharmacy as the concomitant prescription of 5 or more drugs in an individual (15). Here, polypharmacy was divided into 3 ordered categories: no polypharmacy (1-4 drugs), nonexcessive polypharmacy (5-9 drugs), and excessive polypharmacy ( ≥ 10 drugs) person per month (15, 32, 33).

Pharmaceutical spending: Data for spending in 2018 and prior years come from the NCHIR database, which tracks purchases of medications by each people with “Bimeh Salamt”. Information on drug prescription and pharmaceutical spending was extracted manually from the EMR by a trained research assistant and entered into a standardized and pretested case report form. All costs are expressed in US dollars person per month in 2018 prices and exchange rates.

>Statistical Analyses

Prevalence was estimated by dividing the number of individuals who received drugs from each group during a 12-month period by the 2018 city population (1-year prevalence). Age- and sex-specific prevalence patterns were explored graphically. Age-standardized or age and sex-standardized prevalence figures were obtained by direct standardization to the entire Iran population (2016 Iran Census) when appropriate. For direct standardization, the entire Iran population (2016 Iran Census) were chosen as a reference or standard population. The observed number of cases calculated in the populations of interest. Then, apply the age (sex)-specific prevalences from the chosen reference population to the populations of interest. The number of people in each age (sex) group of the populations of interest multiplied by the age (sex)-specific prevalence in the comparable age (sex) group of the reference population. Sum the total number of expected cases for each population of interest. Finally, divide the total number of observed cases of the populations of interest by the expected cases.

Categorical variables were calculated as frequencies (%), and continuous variables were presented as the mean and SD or median and interquartile range (IQR). The results of the χ2-tests were expressed as P values, and the strength of association in the multilevel ordinal logistic regression analyses was expressed as odds ratios (OR) with their corresponding 95% CIs and P values after adjusting for the province and cities clustering effects. In the multilevel ordinal logistic regression (ML-OLR) models, the point estimates, 95% CIs and P values of the factors in both type of univariate and multiple analysis were calculated. Prevalence as a proportional response was estimated from a multilevel beta regression (ML-BR) model. A 2-tailed P < .05 was considered statistically significant. All analyses were performed using R (Version 4.0.3) and SPSS Version 16.0 (IBM Corp). Maps were drawn in Data wrapper. This study was approved by the Research Ethics Committee of Hamadan University of Medical Sciences (Approval no. 16/35/3506). A waiver of patient consent was granted due to the use of anonymized insurance claims data.

>Multilevel Ordinal Logistic Regression Model

Explanatory models for ordinal response variable collected during a single time frame have been previously reviewed by Agresti (34), Bender and Benner (35), and O’Connell (36). Such work was adapted to fit the needs of a hierarchical context. To mention a few, Fielding et al (37) and O’Connell, and Doucette (38) presented the application of the generalized multilevel ordinal model to educational data using the distribution of the latent variable. However, the multilevel ordinal model is somewhat of underutilized method in clinical and epidemiological research studies.

When data are collected in clustering format, methodologies for the handling of ordinal outcomes need to be combined with methods that address the multilevel nature of hierarchical data. Thus, event history data have a 3-level hierarchical structure with responses (level 1) nested within cities (level 2) again nested within provinces (level 3). In the current work, a three-level ordinal analysis is applied.

Suppose ordered values, k = 1,2,…,K is assigned to a latent variable Y related with level one unit k nested within level 2 unit j nested within level 3 unit i. The level 3 units consist of patients’ characteristics, while levels 2 and 1 units of hierarchically measured factors. The multilevel ordinal logistic models’ cumulative probabilities of the response variable (ie, polypharmacy categories) rather than category probabilities using the logit link function are as follows:

(1)

Where k = 1,2,…, K-1; δ(k) are the K-1 intercept terms to model the marginal frequencies in the K ordered categories X' a known matrix associated with the fixed effect β and c and p are random effects that are assumed to follow a multivariate normal distribution with mean zero vector and variance. Σ Σ

Multilevel Beta Regression Model

Let Y be the prevalence of city j=1,2,…,n within province i=1,2,…,I . The prevalence such that Y is a random variable defined in the interval (0, 1). In particular, we assume

follows a beta distribution with mean μ that represents the prevalence of city j within province i, and ϕ is a precision parameter, and the greater its value the lesser the variance of Y, assuming ϕ to be constant. In what follows, we describe the proposed modelling of the μ.

Let X. be a p-dimensional vector of covariates, and β a p-dimensional vector of coefficients, with x and β = (β. Also, let Z be a q-dimensional vector of covariates with , b and a q-dimensional vector of random effects. The multilevel beta regression model is as follow:

(2)

The model specification is completed by assuming Gaussian random effects.

Results

The study sample comprised 3,039,629 elderly men and women that nested in 429 cities that nested again within 31 provinces of Iran, whose characteristics are summarized in Table 1.. The mean patient age was 73.69 years (SD, 7.12). Among the patients, 48.4% (n = 1,471,180) were men, 36.4% (n = 1,106,425) were between 65-69 years, and 26.3% were prescribed at least 1 drug during the winter season. The mean number of drug prescriptions per patient was 1.46 (SD, 0.81) per month; the median was one (IQR, 1). The mean number of prescribed drugs per patient was 4.32 (SD, 3.04) per month; the median was 4 (IQR, 3.8). Among all elderlies, including those with or without any medications, 57.1% (n = 1,147,439) were prescribed 5 or more drugs during 2018 year.

Table 1.

Characteristics of older adults aged ≥ 65 years who were prescribed any drug between April 2018 to March 2019 in Iran, (n = 3,039,629)

Characteristics		n	% or mean (SD)
Sex	Female	1,568,448	51.60%
	Male	1,471,181	48.40%
Age, years	Median = 72.0, IQR = 10.0		73.69 (7.12)
Age group	65 - 69 years	1,106,534	36.40%
	70 - 74 years	755,388	24.90%
	75 - 79 years	513,889	16.90%
	80 - 84 years	378,174	12.40%
	85+years	285,644	9.40%
Season	Spring	724,774	23.80%
	Summer	741,462	24.40%
	Autumn	774,564	25.50%
	Winter	798,829	26.30%
Chronic disease (NCDs)	No	653,520	21.50%
	Yes	2,386,109	78.50%
No. drug prescription, per month	Median = 1.0, IQR = 1.0		1.46 (0.81)
1prescription	2,077,410	68.30%
2 - 3 prescriptions	873,753	28.70%
> 4prescriptions	88,466	2.90%
No. prescribed drug, per month	Median = 4.0, IQR = 3.8		4.32 (3.04)
no polypharmacy	1 - 4 drugs	1,892,220	62.30%
non-excessive polypharmacy	5 - 9 drugs	965,861	31.70%
excessive polypharmacy	> 10drugs	181,548	6.00%
Total drug cost (US $ per month)	Median = 6.86, IQR = 12.26		19.57 (85.31)
Insurance share (US $ per month)	Median = 4.96, IQR = 9.07		15.74 (72.25)
Insured share (US $ per month)	Median = 1.76, IQR = 3.25		3.83 (15.74)

SD: Standard Deviation; IQR: Interquartile range (Q3 - Q1).

Prescription expenditures related to elderly patients in Iran during 2018 totaled $59.48 million, with the 49.48% ($29.43 million) spending among old men. Drug cost for each elderly patient was $6.86 (IQR, 12.26) per month, with $4.96 and $1.76 the insurance and the insured shares, respectively.

In this study, the number of drug prescriptions was recorded monthly for each elderly patient during 12-month. Based on our findings, in Iran and during 2018, there were 2,295,148 and 2,127,534 drug prescriptions among elderly women and men, respectively. According to mapping results, cities of Tehran, Isfahan, Mashhad, Shiraz, Tabriz, Rasht, Sari, Ahvaz, Kerman, and Karaj were ranked based on the most frequent number of prescriptions in both men and women. The estimated prevalence, for number of drug prescriptions > 4, ranged between almost zero in Bashagard, a city in Hormozgan province, to 28.57% in Ghaleye-ganj, a city in southeast of Kerman province. In addition, we estimated the prevalence of > 4 prescriptions per month by sex in all cities of Iran. Results are shown in Fig. 1. Prevalence of equal to four and more drug prescriptions, person per month, in elderly patients, Iran 2018.

Fig. 1. Prevalence of equal to four and more drug prescriptions, person per month, in elderly patients, Iran 2018

In elderly women, Bashagard (0.01%), Deylam (0.03%), Faryab (0.03%), Fanouj (0.06%), and Meshginshahr (0.08%) were 5 cities with the lowest estimated prevalence of > 4 prescriptions per month; and Malekshahi (15.56%), Khamir (15%), Rabar (12.73%), Fahraj (11.11%), and Faroj (10.39%) were 5 cities with the highest estimated prevalence of > 4 prescriptions per month (left side of Fig. 1. Prevalence of equal to four and more drug prescriptions, person per month, in elderly patients, Iran 2018). Furthermore, in elderly men, Meshginshahr (0.01%), Eshtehard (0.02%), Dalgan (0.02%), Normashir (0.04%), and Bashagard (0.05%) were 5 cities with the lowest estimated prevalence of > 4 prescriptions per month; and Ghaleye-ganj (28.57%), Hamoun (11.10%), Bahmai (8.96%), Galikash (7.63%), and Hamidieh (7.14%) were 5 cities with the highest estimated prevalence of > 4 prescriptions per month (right side of Fig. 1. Prevalence of equal to four and more drug prescriptions, person per month, in elderly patients, Iran 2018).

Elderly people in Iran have taken about 13.1 million prescribed drugs during 1 year, which 52.8% (n = 6,938,528) of drug counts related to elderly women. The number of drugs used also decreased with age ( Appendix A1 & Appendix A2 ). More than 13% of prescriptions had 1 drug, and around 37.7% had more than 4 drugs. The maximum of drugs in a prescription was 24.

The estimated number of drug use varied considerably between regions. According to mapping results, cities of Tehran (14.3%), Isfahan (10.1%), Mashhad (5.9%), Shiraz (5.2%), Tabriz (3.2%), Rasht (2.3%), Sari (2.3%), Ahvaz (1.5%), Kerman (1.5%), and Karaj (1.4%) were ranked based on the most frequent number of prescribed drugs in both men and women. Estimated prevalence for number of excessive polypharmacy ( ≥ 10 drugs) ranged between almost zero in Arzuiyeh, a city in Kerman province, to 25% in Kavar, a city in southeast of Fars province. We estimated the prevalence of excessive polypharmacy ( ≥ 10 drugs) per month by sex in all cities of Iran. Results are presented in Fig. 2. Prevalence of excessive polypharmacy ( ≥ 10 drugs), person per month, in elderly patients, Iran 2018.

Fig. 2. Prevalence of excessive polypharmacy ( ≥ 10 drugs), person per month, in elderly patients, Iran 2018

In elderly women, Arzuiyeh (0.01%), Eshtehard (0.01%), Anar (0.02%), Andika (0.03%), and Bastak (0.04%) were 5 cities with the lowest estimated prevalence of excessive polypharmacy per month; and Kavar (25%), Malekshahi (21.74%), Galikash (16.51%), Zirkouh (15.79%), and Rabar (15.38%) were 5 cities with the highest estimated prevalence of excessive polypharmacy per month (left side of Fig. 2. Prevalence of excessive polypharmacy ( ≥ 10 drugs), person per month, in elderly patients, Iran 2018). Furthermore, in elderly men, Arzuiyeh (0.01%), Eshtehard (0.02%), Anar (0.02%), Andika (.03%), and Eejrud (0.04%) were 5 cities with the lowest estimated prevalence of excessive polypharmacy per month; and Hamidieh (22.22%), Kolaleh (14.98%), Galikash (12.75%), Ghaleye-Ganj (12.50%), and Dashti (12.50%) were 5 cities with the highest estimated prevalence of excessive polypharmacy per month (right side of Fig. 2. Prevalence of excessive polypharmacy ( ≥ 10 drugs), person per month, in elderly patients, Iran 2018).

Expenditures on prescribed drugs amounted to $11.63 million ($3.83 per capita) with more contributed spending ($5.89 million) among elderly women in Iran in 2018, excluding insurance rebates (see Table 1. and Table 2.). In per capita terms, the top province spenders were Tehran ($5.06), Azarbayjan-e-sharqi ($4.62), and Golestan ($4.30), with over $3.68 million, and the lowest province spenders (with complete data) were Fars ($2.29), Mazandaran ($2.34), Ilam ($2.56), and Kohgiluyeh va bowyer ahmad ($2.68), with less than $1.1 thousand during 2018.

Table 2.

Associations of patient characteristics with no polypharmacy (1-4 drugs), non-excessive polypharmacy (5-9 drugs) and excessive polypharmacy ( ≥ 10 drugs) among older patients aged ≥ 65 years

Variable		No. concomitant drugs, n (%) or median (IQR)			P-value
		1 - 4	5 - 9	> 10
Sex	Female	951,535 (60.7%)	516,461 (32.9%)	100,452 (6.4%)	< 0.001
	Male	940,685 (63.9%)	449,400 (30.5%)	81,096 (5.5%)
Age group (years)	65 - 69	710,866 (64.2%)	334,993 (30.3%)	60,675 (5.5%)	< 0.001
	70 - 74	469,428 (62.1%)	240,513 (31.8%)	45,447 (6.0%)
	75 - 79	312,090 (60.7%)	169,257 (32.9%)	32,542 (6.3%)
	80 - 84	226,730 (60.0%)	126,491 (33.4%)	24,953 (6.6%)
	85+	173,106 (60.6%)	94,607 (33.1%)	17,931 (6.3%)
Season	Spring	429,677 (59.3%)	243,880 (33.6%)	51,217 (7.1%)	< 0.001
	Summer	467,627 (63.1%)	232,227 (31.3%)	41,608 (5.6%)
	Autumn	490,124 (63.3%)	241,117 (31.1%)	43,323 (5.6%)
	Winter	504,792 (63.2%)	248,637 (31.1%)	45,400 (5.7%)
Chronic disease (NCDs)	No	409,103 (62.6%)	218,929 (33.5%)	25,488 (3.9%)	< 0.001
	Yes	1,379,171 (57.8%)	816,049 (34.2%)	190,889 (8.0%)
No. prescription (per month)	1	1,648,458 (79.4%)	424,444 (20.4%)	4,508 (0.2%)	< 0.001
	2 - 3	243,133 (27.8%)	515,371 (59.0%)	115,249 (13.2%)
	> 4	629 (0.7%)	26,046 (29.4%)	61,791 (69.8%)
Total drug cost (US $ per month)		4.08 (7.19)	11.53 (12.79)	22.74 (30.61)	< 0.001
Insurance share (US $ per month)		2.95 (5.34)	8.29 (9.55)	16.41 (23.39)	< 0.001
Insured share (US $ per month)		1.04 (1.94)	3.16 (3.72)	6.20 (6.84)	< 0.001

Total n = 3,039,629; IQR: Interquartile range (Q3 - Q1). P-value conducted from Chi-square (c2) test.

In elderly women, Sirik ($0.07), Ghasr-e-gand ($0.1), Normashir ($0.24), Karoun ($0.28), and Doureh ($0.45) were 5 cities with the lowest monthly per capita spenders; and Farsan ($7.52), Bandar-e-gaz ($6.47), Rezvanshahr ($6.11), Rabar ($5.64), and Talesh ($5.63) were 5 cities with the highest monthly per capita spenders (left side of Fig. 3. Mean of pharmaceutical spending, USD $ person per month, in elderly patients, Iran 2018). Furthermore, in elderly men, Bashagard ($0.04), Fanouj ($0.06), Kavar ($0.07), Faryab ($0.11), and Jask ($0.24) were 5 cities with the lowest monthly per capita spenders; and Haftgol ($16.32), Bavi ($14.91), Kalat ($10.96), Razo jalgelan ($10.96), and Mobarakeh ($8.18) were 5 cities with the highest monthly per capita spenders in Iran 2018 (right side of Fig. 3. Mean of pharmaceutical spending, USD $ person per month, in elderly patients, Iran 2018).

Fig. 3. Mean of pharmaceutical spending, USD $ person per month, in elderly patients, Iran 2018

Tables 2 and 3 show the results of the univariate analyses with chi-square tests and multivariate analyses with multilevel ordinal logistic regressions, respectively.

Table 3.

Factors associated with no monthly polypharmacy (1-4 drugs), non-excessive polypharmacy (5-9 drugs), and excessive polypharmacy ( ≥ 10 drugs) among elderly patients aged ≥65 years (multilevel ordinal logistic regression analysis results)

Multivariable analysis				Univariate analysis
Parameter		OR	95% CI		P-value	OR	95% CI		P-value
Lower	Upper	Lower	Upper
Age group	65 - 69	0.841	0.812	0.870	< 0.001	0.857	0.828	0.886	< 0.001
	70 - 74	0.905	0.873	0.938	< 0.001	0.933	0.901	0.967	< 0.001
	75 - 79	0.950	0.915	0.987	0.009	0.981	0.945	1.019	0.328
	80 - 84	1.007	0.967	1.049	0.726	1.023	0.983	1.064	0.265
	85+¶	1	-	-	-	1	-	-	-
Sex	Female	1.164	1.142	1.186	< 0.001	1.134	1.113	1.156	< 0.001
	Male¶	1	-	-	-	1	-	-	-
Season	Spring	1.274	1.241	1.309	< 0.001	1.210	1.179	1.242	< 0.001
	Summer	1.029	1.002	1.057	0.038	0.996	0.970	1.023	0.791
	Autumn	1.005	.979	1.033	0.692	0.982	0.957	1.008	0.181
	Winter¶	1	-	-	-	1	-	-	-
Chronic disease (NCDs)	Yes	2.174	2.069	2.275	< 0.001	2.149	2.082	2.188	< 0.001
	No¶	1	-	-	-	1	-	-	-
Total cost $		1.163	1.159	1.168	< 0.001	1.007	1.007	1.008	< 0.001
Insurance share $		1.009	1.005	1.013	< 0.001	1.006	1.006	1.006	< 0.001
Insured share $		1.053	1.048	1.057	< 0.001	1.087	1.085	1.090	< 0.001
Variance component from multivariable analysis
Estimate		Standard Error			P-value
Sigma (Province)		0.266		0.040			< 0.001
Sigma (City)		1.461		0.088			< 0.001

-2 log-Likelihood = 268305.044, AIC = 268129.029; BIC = 268250.459 OR: Odds Ratio; CI: Confidence Interval; ¶: Reference level.

The distribution of polypharmacy (No. concomitant drugs) by variables are shown in Table 2.. At the all level of the response (1 to 3), polypharmacy is more frequent (percent) in elderly women, younger age group, spring season, and elderlies with at least 1 chronic condition (NCDs). However, in relation to number of prescriptions, nonexcessive polypharmacy in level 2 (2-3 prescriptions per month) and excessive polypharmacy in level 3 (>4 prescriptions per month) are more frequent.

The multivariate analyses identified the following factors as being significantly associated with ordered response, in order: 1 = no polypharmacy (1-4 drugs), 2 = nonexcessive polypharmacy (5-9 drugs), and three = excessive polypharmacy ( ≥ 10 drugs): age group (reference: 85+ years), sex (reference: male), season (reference: winter), chronic disease (reference: no), drug cost, insurance share, and insured share (all P < 0.001).

As can be seen, the variance components of random effects were significant in provinces and cities for the proposed model (multilevel ordinal logistic regression) (Table 3. ). Therefore, it can be concluded that there was a significant heterogeneity among cities and provinces in terms of the studied response variable. Therefore, all estimates reported and interpreted after adjusting for the province and cities clustering effects.

With increasing age, the chance of higher level of polypharmacy decreased also, so that the odds of higher level of polypharmacy for people in the age group 85+ years are 1.19 (OR, 1/0.841 = 1.19) greater than people in age group 65-69 years, given that all of the other variables in the model are held constant. In addition, odds of higher level of polypharmacy for people in the age group 85+ years are 1.10 (OR, 1/0.841 = 1.10) greater than people in age group 70-74 years, showing a statistically significant effect, P < 0.001.

For elderly women, the odds of polypharmacy (excessive and nonexcessive vs no polypharmacy) were 1.164 (95% CI, 1.142-1.186) times that of elderly men. In addition, in the spring, the odds of polypharmacy were 1.274 (95% CI, 1.241-1.309) times that of winter, indicating a statistically significant effect, P < 0.001.

Similarly, polypharmacy was strongly higher among patients who had a disease of NCDs (OR, 2.174; 95% CI, 2.069-2.275 (P < 0.001)).

In pharmaceutical spending characteristics, an increase in factors of the total drug cost (OR, 1.163; 95% CI, 1.159-1.168), insurance share (OR, 1.009; 95% CI, 1.005-1.009), and insured share (OR, 1.053; 95% CI, 1.048-1.057) also significantly associated with an increase in the odds of higher level of polypharmacy (all P < 0.001).

Discussion

The present study mapped and elucidated the drug prescriptions, polypharmacy, and pharmaceutical spending and the factors associated with polypharmacy in >3 million patients aged ≥65 years residing in 429 cities in Iran.

Approximately 4.4 million with a mean of 1.46 per capita drug prescriptions have been distributed to elderly patients during 2018 in Iran. The value of this index is about one-third of the results of other studies. According to data available in the United States, 16% of adults are over 60 years and account for 40% of the prescribed medications, with a mean of 4.5 medicines per person (39).

Approximately, 38% of participants were receiving polypharmacy, showing that polypharmacy in the elderly is currently the norm and not the exception. Concomitant use of multiple prescription drugs (‘polypharmacy‘) is increasingly common, with 10% of the population (39,40) and 30% of older adults in the United States taking 5 or more drugs simultaneously (39-41). A cross-sectional analysis of the Survey of Health, Aging, and Retirement in Europe (SHARE) database showed that the prevalence of polypharmacy, defined as taking 5 or more medications concurrently in older adults aged 65 years or more, was between 26.3% and 39.9% among European countries and Israel (42).

Similar studies using prescribed drug registry data or drug claims data have reported that 44% of Swedes aged ≥70 years (15), 66% of Canadians aged ≥65 years (43), and 77.5% of Qatari elderly patients aged ≥65 years (44) were prescribed 5 or more drugs. It has also been reported that 86% of Koreans aged ≥65 years were prescribed 6 or more drugs (24). Although our estimated prevalence of polypharmacy was between those reported in the European countries and Swedish studies, inherent differences in the definitions of polypharmacy and study settings (eg, data collection methods, healthcare systems, and available drugs) make it difficult to compare results among countries.

The present study showed that 6.5% of all patients aged ≥65 years who were prescribed any drug during the study period were prescribed 10 or more drug types.

The mean cost per prescription was 19.57 US dollars. This was 7.04 US dollars in 2010 (45) and 8.21 US dollars in 2011 (46), which indicated a 3-fold increase in the cost of medicines over the last 5 years. This could be the result of sanctions and a leap in inflation between 2011 and 2018.

Consistent with other studies, the prevalence of polypharmacy in the present study was significantly higher among women (18, 20, 26, 47). In contrast, other studies have reported higher polypharmacy rates in men (17,24). Such contradictions among study findings could be explained by differences in physicians’ prescribing approaches toward genders as well as to differences between genders and their health-seeking behaviors. Consistent with another study in Iran (48,49), the prevalence of almost all of the NCDs we studied, except for cardiovascular diseases, was higher in women than in men. In addition, polypharmacy showed a stronger association with certain NCDs than others. This is consistent with findings from other studies (17, 21, 22, 24, 27).

A study in Norway (2015) that examined the regular general practitioner role in polypharmacy reported that the risk of polypharmacy in patients increases significantly with the number of prescribers (OR, 2.32; 95% CI, 2.31-2.33) (50). One other factor that could explain such a phenomenon is the high prevalence of NCDs in our study population. Furthermore, a study in Canada showed a significant association between polypharmacy and higher frequency of family physician visits in elderly patients (51). Finally, the provision of medications at lower or no cost to elderly Iranian citizens might make it easier for physicians to prescribe them more.

Finally, in our study, polypharmacy was substantially increased with the increasing insurance share in spending costs. This is consistent with findings from other studies in Sweden (16,52).

The present study is the first to map the prevalence of polypharmacy and its associated factors among elderly citizens in Iran. The strength of our study lies in its large sample size, which allowed for statistical analysis with sufficient statistical power. Therefore, the findings from this study provide a reliable basis to confirm that high polypharmacy exists in elderly patients in the cities and provinces of Iran. Moreover, the use of standardized estimates, such as the prevalence of polypharmacy and prescription, made the results more valid, reliable, and enabled comparison with other studies.

On the other hand, the lack of a standard definition of number of prescription and polypharmacy across studies made comparisons difficult. Moreover, the calculated prevalence might be overestimated because elderly patients who we extracted might have suffered from multiple health conditions.

Another limitation is that some of the variables (eg, socioeconomic status, marital status, body mass index, or exact NCDs disease) were not consistently recorded for most patients, making us unable to include them in our analysis.

Finally, because of our study design, our findings cannot be generalized to the entire population of interest, but can be only applied to the population included in the study.

Conclusion

This study provided evidence that the prevalence of polypharmacy among Iranian elderly patients is very high, with almost 3 quarters of the study population exposed to it. The study as well demonstrated a significant association between polypharmacy and older age with about 42% of the study participants being older than 75 years. Our findings confirmed the strong relationship between polypharmacy and NCDs, such as hypertension, diabetes mellitus, dyslipidemia, cardiovascular diseases, and asthma. As appropriate care for elderly patients is increasingly challenging, targeted educational programs should be developed for health care professionals to raise their awareness of the magnitude and negative impact of polypharmacy. Furthermore, primary health care centers should establish best practice guidelines for improved medical practice in the prescription of medications for such a vulnerable population.

Conflict of Interests

The authors declare that they have no competing interests.