Self-medication before and during the COVID-19 pandemic in Türkiye: evidence from national health surveys

Abstract

Background

Sociodemographic characteristics, health status, and access to health care are important determinants of health behaviors and may shift during periods of health system disruption such as the COVID-19 pandemic. This study aimed to examine whether these determinants were associated with self-medication and whether the prevalence of self-medication changed in Türkiye before and during the COVID-19 pandemic.

Methods

This study used data from the Türkiye Health Survey conducted in 2019 (n = 16,253) and 2022 (n = 21,444), which was designed to provide national-level estimates for Türkiye. Descriptive analyses and chi-square tests were used to compare the prevalence of self-medication across survey periods. Multivariable modified Poisson regression models with robust variance were estimated separately for the pre-COVID-19 and during-COVID-19 periods to identify factors associated with self-medication and to estimate adjusted prevalence ratios (aPRs) with 95% confidence intervals. To assess whether associations differed between the two periods, pooled modified Poisson regression models were constructed including an indicator for survey period and interaction terms between the COVID-19 period and selected determinants. The statistical significance of interaction effects was evaluated using global Wald chi-squared tests.

Results

The prevalence of self-medication increased significantly from the pre-COVID-19 period (34.9%) to the during-COVID-19 period (38.8%) (p < 0.001). In pooled analyses, self-medication remained more prevalent during the COVID-19 period (aPR = 1.11, 95% CI: 1.08–1.14). Across both survey periods, higher prevalence of self-medication was observed among women, individuals with higher educational attainment, those reporting moderate or severe bodily pain, individuals with long-standing illness, tobacco or alcohol users, and those experiencing unmet health care needs due to waiting lists. Global interaction tests indicated that the associations of educational attainment, age group, and health insurance coverage with self-medication differed significantly between the pre- and during-COVID-19 periods, whereas the associations with gender, long-standing illness, and self-perceived health remained stable.

Conclusions

Self-medication prevalence was higher during the COVID-19-period survey in Türkiye, and the associations between self-medication and several sociodemographic and health-related determinants differed between survey periods. These findings suggest that changes in health care access and perceived health needs occurring during public health crises may be associated with differences in self-medication patterns, highlighting the importance of monitoring self-medication behaviors and supporting vulnerable population groups.

Introduction

Self-medication is widely recognized as a component of self-care. The World Health Organization (WHO) defines self-medication as the use of medicines, vitamins, or dietary supplements by individuals for the purpose of treating self-recognized diseases or symptoms in the absence of professional medical advice [1]. Before the COVID-19 pandemic, the prevalence of self-medication varied substantially across populations, with reported estimates ranging from 11.2% to 93.7% depending on geographical setting, study population, and methodological differences across studies [2]. Such variability may also reflect temporal changes in health-seeking behavior, including increasing access to the Internet and digital health information, which may facilitate self-medication practices. Disruptions in health care access during the COVID-19 pandemic may have further influenced these patterns. A recent systematic review estimated the pooled global prevalence of self-medication at 48.6%, with notable differences across regions, population groups, and pandemic periods, pointing to an overall increase in self-medication during the COVID-19 pandemic [3]. Although self-medication may support individuals in managing minor health problems, it also raises several public health concerns, such as inappropriate medication use, incorrect dosing, potential drug interactions, and delayed diagnosis of underlying conditions [4].

Türkiye is an upper-middle-income country with a population of approximately 85 million people. The Turkish health system is organized under the Ministry of Health and provides universal health coverage through the General Health Insurance system. Primary health care services are delivered mainly through the Family Physician system, which serves as the first point of contact for most health needs. According to national health statistics, Türkiye has experienced substantial expansion in health service coverage and utilization over the past two decades, with widespread access to primary care and pharmaceutical services across the country. This health system context is important for understanding patterns of health-seeking behavior, including the use of self-medication. In Türkiye, studies conducted in specific population groups, such as university students and other selected samples, have reported prevalence estimates ranging from 26.3% to 89.9%, reflecting differences in study populations and methodologies. However, population-based evidence on self-medication patterns in the general population remains limited [5].

Previous research shows that self-medication is shaped by a combination of sociodemographic characteristics, health-related needs, and access to health care [6, 7]. Sociodemographic characteristics such as gender, age, educational attainment, marital status, and employment status have been frequently associated with self-medication, although findings differ across countries and settings [8–10]. Health-related factors, including perceived health status, bodily pain, chronic conditions, depressive symptoms, and body mass index, further influence individuals’ perceived need for treatment [8, 11]. In addition, barriers to accessing formal health care—such as long waiting times, financial constraints, and limited insurance coverage—have been consistently linked to higher levels of self-medication [6–8]. Taken together, these factors can be conceptualized as reflecting three interrelated domains influencing self-medication behavior: individual sociodemographic characteristics, health-related needs, and access to health care services.

The COVID-19 pandemic substantially disrupted health care systems and altered health-seeking behaviors worldwide. Restrictions on movement, appointment delays, and concerns about exposure to SARS-CoV-2 discouraged many individuals from seeking professional medical care [2, 3]. At the same time, heightened perceptions of risk and uncertainty led individuals to adopt more self-directed approaches to managing symptoms. Social factors, including fear of isolation and stigma associated with a COVID-19 diagnosis, may have further reinforced self-medication practices during this period [12]. As a result, the pandemic-period context may have been associated not only with higher prevalence of self-medication but also with differences in how established determinants were related to this behavior. In this context, disruptions in health care access and changes in perceived health risks may help explain why health-related needs and barriers to care appeared more strongly associated with self-medication practices during the pandemic-period survey.

While a growing body of international literature has examined self-medication practices during the COVID-19 pandemic across Europe, Asia, the Americas, and the Middle East [12–15], population-based evidence on self-medication in the general population of Türkiye remains scarce [5]. Moreover, few studies have formally compared self-medication patterns before and during the pandemic while examining whether the relationships between key determinants and self-medication changed during this period Understanding these potential shifts is important because disruptions in health care access and changes in perceived health needs during the pandemic may have been associated with differences in self-medication behavior. Addressing this gap, the present study uses data from the Türkiye Health Surveys conducted in 2019 and 2022, which were designed to provide national-level estimates for Türkiye. Guided by a conceptual framework that considers sociodemographic characteristics, health-related needs, and access to health care as key determinants of self-medication, the study aims to (i) estimate the change in the prevalence of self-medication before and during the COVID-19 pandemic, (ii) identify factors associated with self-medication, and (iii) assess whether the associations between selected determinants and self-medication differed between the pre- and during-pandemic periods using pooled regression models with interaction terms.

Materials and methods

Data source and study population

This cross-sectional study was based on secondary analysis of data from the 2019 and 2022 Türkiye Health Surveys (THS) conducted by the Turkish Statistical Institute (TurkStat) [16]. The 2019 survey was conducted between September and December 2019, and the 2022 survey between September and December 2022. The latter period corresponded to a later phase of the COVID-19 pandemic in Türkiye, during which most restrictive public-health measures had already been lifted, although broader effects of the pandemic on health behaviors and health care utilization may still have persisted. The survey was designed by TurkStat to produce national-level estimates for Türkiye. It covers the non-institutionalized population residing in households across the country and was carried out through face-to-face interviews. The sampling design was based on a stratified two-stage cluster sampling approach using the National Address Database as the sampling frame; further methodological details are available in the official survey documentation [16]. However, the publicly available microdata did not provide sufficient information regarding sampling weights, primary sampling unit (PSU) identifiers, or stratum identifiers required for fully design-based survey analysis. Therefore, analyses were conducted using conventional regression methods. The 2019 wave of the THS (conducted before the COVID-19 pandemic) included 16,253 individual respondents, while the 2022 wave (conducted during the pandemic) included 21,444 respondents. For the purposes of this study, all participants with complete data on the variables of interest were included. The reporting of the study follows the STROBE guidelines for observational research.

Study measures

The primary outcome was self-medication behavior in the last two weeks, defined as a binary (yes/no) variable based on two survey questions. Participants were asked: (1) “In the last two weeks, have you taken any medications by your own decision or preference (without a doctor’s prescription)?”, and (2) “In the last two weeks, have you taken any dietary supplements or vitamins by your own decision or preference?”. Individuals who answered “yes” to either question were classified as engaging in self-medication. Accordingly, self-medication was operationalized as broad non-prescribed health product use, encompassing medicines, vitamins, and dietary supplements used on one’s own initiative within a two-week recall period. This approach is consistent with broader population-based conceptualizations of self-medication and self-care behaviors reported in previous literatüre [2, 3, 6–8].

A range of sociodemographic, health-related, and health service utilization variables were examined as predictors of self-medication. All variables were based on self-reported responses collected during face-to-face interviews. Sociodemographic variables included gender (male/female), age group (18–29, 30–44, 45–64, ≥ 65 years), educational attainment (no formal education, primary education, secondary education, associate/bachelor’s degree, master’s/doctoral degree), marital status (single, married, divorced, widowed), employment status (employed, unemployed, not in the labor force), and health insurance coverage (yes/no). Health-related variables captured participants’ health status and conditions. These included limitation in daily activities (yes/no), bodily pain (none/very mild/mild, moderate, extreme/severe), self-perceived health status (very good/good, fair, bad/very bad), presence of a long-standing illness or health problem (yes/no), depressive disorder (yes/no), body mass index (underweight, normal, overweight, obese), tobacco use (yes/no), and alcohol use (yes/no).

Health service utilization was assessed using variables reflecting contact with health care services and unmet health care needs. Utilization variables included the frequency of inpatient care utilization in the past year (0, 1–2, 3–5, ≥ 6 hospitalizations), frequency of family physician visits (0, 1–2, 3–5, ≥ 6 visits), and frequency of specialist physician visits (0, 1–2, 3–5, ≥ 6 visits). In addition, unmet health care needs were assessed through two separate indicators: unmet need for health care due to waiting lists (no, yes, no need for health care) and unmet need for health care due to inability to afford care (no, yes, no need for health care).

Statistical analysis

Descriptive statistics were first computed to characterize the 2019 and 2022 samples. Frequencies and percentages were used to describe the distribution of sociodemographic characteristics, health status variables, and healthcare utilization variables, as well as the prevalence of self-medication in each year. Bivariate analyses were then conducted to examine associations between each independent variable and self-medication status. In these analyses, Chi-square (χ²) tests were used to compare proportions between groups – for example, to assess differences in self-medication prevalence across categories of an independent variable and to compare characteristics between those who did and did not self-medicate. Variables considered theoretically relevant based on previous literature and conceptual considerations were subsequently included in the multivariable regression models.

Multivariable regression analyses were conducted to identify factors independently associated with self-medication. Because the prevalence of self-medication in the study population exceeded 10%, modified Poisson regression with a log link and robust variance estimation was used to directly estimate adjusted prevalence ratios (aPRs) and their 95% confidence intervals (CIs). This approach provides more interpretable estimates of association than logistic regression when outcomes are common. Separate multivariable modified Poisson regression models were first fitted for the 2019 and 2022 survey waves to examine factors associated with self-medication within each period. In these models, self-medication (yes/no) was specified as the dependent variable, while sociodemographic characteristics, health-related variables, and indicators of health care utilization were included simultaneously as independent variables (covariates). Adjusted prevalence ratios (aPRs) with 95% confidence intervals were calculated for each predictor. Prior to model estimation, potential multicollinearity among independent variables was assessed using variance inflation factors (VIFs) and generalized variance inflation factors (GVIFs) for multi-category variables. Although several individual dummy-variable VIF values were moderately elevated for some multi-category predictors, all scaled GVIF values remained below commonly used thresholds (maximum GVIF^(1/(2df)) = 1.40), indicating no evidence of problematic multicollinearity at the variable level.

To examine whether the associations between selected determinants and self-medication differed between the pre-COVID-19 and during-COVID-19 periods, an interaction analysis was conducted using a pooled dataset combining the 2019 and 2022 survey waves. A pooled multivariable modified Poisson regression model was fitted including an indicator variable for survey period (2019 vs. 2022), all covariates listed above, and interaction terms between the COVID-19 period and selected independent variables.

The pooled modified Poisson regression model with interaction terms can be expressed as:

where Y denotes self-medication status, period represents the survey period (pre- vs. during-COVID-19), and X represents the independent variables included in the model. The model was estimated using a Poisson working likelihood with a log link function, and variance estimates were obtained using the Huber–White robust sandwich estimator. Interaction terms were specified a priori for a limited set of variables: gender, educational attainment, age group, health insurance coverage, long-standing illness or health problem, and self-perceived health status. These variables were selected because they represent broad sociodemographic and health-related determinants that may plausibly differ in their associations with self-medication across survey periods. Interaction analyses were limited to selected sociodemographic and health-related variables considered most theoretically relevant to potential differences in self-medication patterns across survey periods. This approach was intended to maintain model parsimony, reduce excessive multiple interaction testing, and avoid overparameterization while focusing on conceptually plausible effect modification by the COVID-19 period. Interaction effects were evaluated using global Wald tests for each interaction block. A statistically significant global test indicated that the association between the corresponding predictor and self-medication differed between the pre- and during-COVID-19 survey periods. Interaction effects were evaluated using global Wald tests for each interaction block. A statistically significant global test indicated that the association between the corresponding predictor and self-medication differed between the pre- and during-COVID-19 periods.

Results of the regression analyses are presented as aPRs with 95% confidence intervals. A p-value < 0.05 was considered statistically significant. All statistical analyses were performed in Python. Modified Poisson regression models were estimated using generalized linear models implemented in the statsmodels package, specifying a Poisson family with a log link function. Robust standard errors were obtained using the Huber–White sandwich covariance estimator. Adjusted prevalence ratios were calculated by exponentiating the regression coefficients, and 95% confidence intervals were derived from the robust standard errors.

Results

Table 1 presents the distribution of self-medication prevalence across sociodemographic, health-related, and health care utilization characteristics in the pre- and during-COVID-19 periods. The study included 16,253 participants in the pre-COVID-19 period (2019) and 21,444 participants in the during-COVID-19 period (2022). As shown in Table 1, the prevalence of self-medication increased significantly from the pre-COVID-19 period (34.9%) to the during-COVID-19 period (38.8%) (p < 0.001). In both periods, self-medication was more prevalent among women, individuals with higher educational attainment, and those aged 30–44 years. Higher prevalence rates were also consistently observed among participants who were employed, reported bodily pain, perceived their health as fair, had a long-standing illness or depressive disorder, or used tobacco or alcohol.

Table 1.

Prevalence of self-medication across sociodemographic and health-related characteristics in the pre- and during-COVID-19 periods

		Pre-COVID-19 period (n = 16253)					During-COVID-19 period (n = 21444)
		N		%			N		%		p (χ2)
Self-medication	Yes	5669		34.9			8320		38.8		< 0.001
	No	10,584		65.1			13,124		61.2

		Self-medication				p (χ2)	Self-medication				p (χ2)
		Yes		No			Yes		No
		N	%	N	%		N	%	N	%
Gender	Female	3414	60.2	5460	51.6	< 0.001	4748	57.1	6413	48.9	< 0.001
	Male	2255	39.8	5124	48.4		3572	42.9	6711	51.1
Educational attainment	No formal education	663	11.7	1512	14.3	< 0.001	809	9.7	1525	11.6	< 0.001
	Primary education	2704	47.7	5091	48.1		3566	42.9	5956	45.4
	Secondary education	1107	19.5	2109	19.9		1851	22.2	3128	23.8
	Associate/Bachelor’s degree	1042	18.4	1700	16.1		1847	22.2	2306	17.6
	Master’s/Doctoral degree	153	2.7	172	1.6		247	3.0	209	1.6
Age	18–29	1054	18.6	2307	21.8	< 0.001	1519	18.3	2975	22.7	< 0.001
	30–44	1978	34.9	3025	28.6		2766	33.2	3972	30.3
	45–64	1890	33.3	3541	33.5		2862	34.4	4206	32.1
	≥ 65	747	13.2	1711	16.2		1173	14.1	1971	15.0
Marital status	Single	861	15.2	1932	18.3	< 0.001	1418	17.0	2771	21.1	< 0.001
	Married	4208	74.2	7504	70.9		5969	71.7	9060	69.1
	Divorced	227	4.0	347	3.3		404	4.9	488	3.7
	Widowed	373	6.6	801	7.6		529	6.4	805	6.1
Employment status	Employed	2336	41.2	4132	39.0	0.013	3604	43.3	5406	41.2	< 0.001
	Unemployed	313	5.5	656	6.2		424	5.1	850	6.5
	Not in the labor force	3020	53.3	5796	54.8		4292	51.6	6868	52.3
Health insurance coverage	Yes	5262	92.8	9703	91.7	0.010	7690	92.4	12,215	93.1	0.074
	No	407	7.2	881	8.3		630	7.6	909	6.9
Limitation in daily activities	No	3686	65.0	6963	65.8	0.326	5412	65.0	9498	72.4	< 0.001
	Yes	1983	35.0	3621	34.2		2908	35.0	3626	27.6
Bodily pain	None/Very mild/Mild	3737	65.9	7751	73.2	< 0.001	5611	67.4	10,361	79.0	< 0.001
	Moderate	1042	18.4	1514	14.3		1566	18.8	1678	12.8
	Extreme/Severe	890	15.7	1319	12.5		1143	13.7	1085	8.3
Self-perceived health	Very good/Good	3062	54.0	6178	58.4	< 0.001	4659	56.0	8410	64.1	< 0.001
	Fair	1980	34.9	3154	29.8		2883	34.6	3653	27.8
	Very bad/Bad	627	11.1	1252	11.8		778	9.4	1061	8.1
Long-standing illness or health problem	No	1790	31.6	4310	40.7	< 0.001	3050	36.7	6543	49.9	< 0.001
	Yes	3879	68.4	6274	59.3		5270	63.3	6581	50.1
Depressive disorder	No	5205	91.8	9971	94.2	< 0.001	7833	94.2	12,629	96.2	< 0.001
	Yes	464	8.2	613	5.8		487	5.8	495	3.8
Body mass index (BMI)	Underweight	151	2.7	272	2.6	0.416	232	2.8	310	2.4	< 0.001
	Normal	2069	36.5	3927	37.1		2988	35.9	5063	38.6
	Overweight	2085	36.8	3957	37.4		3175	38.2	4979	38.0
	Obese	1364	24.1	2428	22.9		1925	23.1	2772	21.1
Tobacco use	No	3726	65.7	7375	69.7	< 0.001	5333	64.1	8956	68.2	< 0.001
	Yes	1943	34.3	3209	30.3		2987	35.9	4168	31.8
Alcohol use	No	5276	93.1	9987	94.4	< 0.001	7720	92.8	12,497	95.2	< 0.001
	Yes	393	6.9	597	5.6		600	7.2	627	4.8
Inpatient care utilization	0	4993	88.1	9337	88.2	0.031	7409	89.1	11,864	90.4	0.003
	1–2	367	6.5	590	5.6		493	5.9	630	4.8
	3–5	124	2.2	274	2.6		167	2.0	242	1.8
	≥ 6	185	3.3	383	3.6		251	3.0	388	3.0
Family physician visits	0	3485	61.5	6798	64.2	< 0.001	4710	56.6	7784	59.3	< 0.001
	1–2	1661	29.3	3005	28.4		2680	32.2	3872	29.5
	3–5	392	6.9	612	5.8		732	8.8	1199	9.1
	≥ 6	131	2.3	169	1.6		198	2.4	269	2.0
Specialist physician visits	0	3382	59.7	6564	62.0	< 0.001	5058	60.8	8610	65.6	< 0.001
	1–2	1599	28.2	2982	28.2		2300	27.6	3186	24.3
	3–5	484	8.5	776	7.3		730	8.8	1050	8.0
	≥ 6	204	3.6	262	2.5		232	2.8	278	2.1
Unmet need for health care due to waiting lists	No	3859	68.1	7800	73.7	< 0.001	4579	55.0	8429	64.2	< 0.001
	Yes	1667	29.4	2391	22.6		3612	43.4	4328	33.0
	No need for health care	143	2.5	393	3.7		129	1.6	367	2.8
Unmet need for health due to inability to afford	No	4778	84.3	9181	86.8	< 0.001	6913	83.1	11,018	83.9	< 0.001
	Yes	627	11.1	801	7.6		829	10.0	952	7.3
	No need for health care	264	4.7	602	5.7		578	6.9	1154	8.8

Statistically significant p-values (p < 0.05) are shown in bold

In the during-COVID-19 period, self-medication prevalence increased across almost all sociodemographic and health-related subgroups, with particularly pronounced differences among individuals reporting limitations in daily activities, higher health care utilization, and unmet health care needs due to waiting lists or inability to afford care (all p < 0.001).

Table 2 presents the results of modified Poisson regression analyses with robust variance examining factors associated with self-medication in the pre- and during-COVID-19 periods. In both periods, female gender was independently associated with a higher prevalence of self-medication compared with males. Educational attainment showed a clear dose–response pattern, with progressively higher prevalence ratios observed among individuals with increasing levels of education. This gradient was more pronounced in the during-COVID-19 period, particularly among individuals with tertiary education. Age-related patterns differed between the two periods. In the pre-COVID-19 period, individuals aged 30–44 years had a higher prevalence of self-medication compared with those aged ≥ 65 years, whereas no significant age-related differences were observed in the during-COVID-19 period. Marital status was modestly associated with self-medication. Married individuals had a slightly higher prevalence of self-medication than single individuals in both periods, while divorced individuals showed a higher prevalence only in the during-COVID-19 period. Widowed status was not significantly associated with self-medication. Employment status showed period-specific differences. Being employed was not significantly associated with self-medication before the pandemic but was associated with a higher prevalence of self-medication in the during-COVID-19 period. Lack of health insurance coverage was not associated with self-medication in the pre-COVID-19 period but became a significant factor in the during-COVID-19 period.

Table 2.

Adjusted prevalence ratios (aPRs) for factors associated with self-medication in the pre-COVID-19 and during-COVID-19 periods estimated using modified Poisson regression with robust variance

Variable		Pre-COVID-19 period		During-COVID-19 period
		aPR (95% CI)	p	aPR (95% CI)	p
Gender	Male	reference		reference
	Female	1.31 (1.24–1.38)	< 0.001	1.26 (1.21–1.31)	< 0.001
Educational attainment	No formal education	reference		reference
	Primary education	1.17 (1.08–1.26)	< 0.001	1.20 (1.13–1.28)	< 0.001
	Secondary education	1.22 (1.12–1.33)	< 0.001	1.33 (1.24–1.44)	< 0.001
	Associate/Bachelor’s degree	1.32 (1.21–1.45)	< 0.001	1.57 (1.46–1.69)	< 0.001
	Master’s/Doctoral degree	1.67 (1.46–1.92)	< 0.001	1.87 (1.68–2.09)	< 0.001
Age	18–29	1.01 (0.92–1.12)	0.781	0.93 (0.86–1.01)	0.090
	30–44	1.12 (1.03–1.23)	0.007	1.01 (0.94–1.08)	0.809
	45–64	1.01 (0.94–1.10)	0.729	1.01 (0.95–1.07)	0.856
	≥ 65	reference		reference
Marital status	Single	reference		reference
	Married	1.09 (1.01–1.17)	0.024	1.08 (1.02–1.14)	0.009
	Divorced	1.06 (0.93–1.20)	0.387	1.10 (1.00–1.20)	0.043
	Widowed	1.00 (0.88–1.13)	0.975	1.07 (0.98–1.18)	0.149
Employment status	Unemployed	reference		reference
	Employed	1.06 (0.96–1.17)	0.228	1.13 (1.04–1.23)	0.003
	Not in the labor force	0.96 (0.87–1.07)	0.492	1.04 (0.95–1.13)	0.416
Health insurance coverage	Yes	reference		reference
	No	0.93 (0.86–1.02)	0.111	1.15 (1.08–1.22)	< 0.001
Limitation in daily activities	No	reference		reference
	Yes	0.85 (0.80–0.90)	< 0.001	1.01 (0.97–1.06)	0.667
Bodily pain	None/Very mild/Mild	reference		reference
	Moderate	1.17 (1.11–1.24)	< 0.001	1.23 (1.18–1.29)	< 0.001
	Extreme/Severe	1.18 (1.10–1.26)	< 0.001	1.32 (1.25–1.39)	< 0.001
Self-perceived health	Very good/Good	reference		reference
	Fair	1.07 (1.01–1.13)	0.021	1.02 (0.97–1.06)	0.510
	Very bad/Bad	0.97 (0.88–1.06)	0.478	0.95 (0.88–1.02)	0.185
Long-standing illness or health problem	No	reference		reference
	Yes	1.30 (1.23–1.38)	< 0.001	1.28 (1.22–1.34)	< 0.001
Depressive disorder	No	reference		reference
	Yes	1.13 (1.05–1.22)	0.001	1.05 (0.98–1.13)	0.152
Body mass index (BMI)	Underweight	reference		reference
	Normal	0.98 (0.86–1.12)	0.818	0.89 (0.81–0.98)	0.016
	Overweight	0.99 (0.86–1.13)	0.826	0.92 (0.84–1.02)	0.107
	Obese	1.00 (0.88–1.15)	0.950	0.94 (0.84–1.04)	0.198
Tobacco use	No	reference		reference
	Yes	1.14 (1.09–1.19)	< 0.001	1.15 (1.11–1.19)	< 0.001
Alcohol use	No	reference		reference
	Yes	1.17 (1.08–1.26)	< 0.001	1.23 (1.16–1.31)	< 0.001
Inpatient care utilization	0	reference		reference
	1–2	1.01 (0.93–1.10)	0.752	1.05 (0.98–1.13)	0.137
	3–5	0.86 (0.74–0.99)	0.035	0.97 (0.86–1.09)	0.601
	≥ 6	0.96 (0.85–1.08)	0.500	0.94 (0.85–1.04)	0.216
Family physician visits	0	reference		reference
	1–2	1.02 (0.97–1.07)	0.380	0.99 (0.96–1.03)	0.742
	3–5	1.11 (1.02–1.21)	0.021	0.93 (0.87–0.99)	0.022
	≥ 6	1.20 (1.04–1.38)	0.013	1.01 (0.90–1.14)	0.826
Specialist physician visits	0	reference		reference
	1–2	0.95 (0.90–1.00)	0.037	1.02 (0.98–1.06)	0.353
	3–5	1.03 (0.95–1.11)	0.538	1.01 (0.95–1.08)	0.760
	≥ 6	1.13 (1.00–1.27)	0.042	1.05 (0.95–1.17)	0.352
Unmet need for health care due to waiting lists	No	reference		reference
	Yes	1.12 (1.07–1.17)	< 0.001	1.16 (1.12–1.20)	< 0.001
	No need for health care	0.88 (0.74–1.05)	0.160	0.87 (0.75–1.02)	0.094
Unmet need for health due to inability to afford	No	reference		reference
	Yes	1.16 (1.08–1.24)	< 0.001	1.03 (0.97–1.09)	0.327
	No need for health care	1.03 (0.91–1.17)	0.660	0.95 (0.89–1.02)	0.180

Abbreviations: aPR adjusted prevalence ratio, CI confidence interval

Both models were adjusted for all variables listed. Reference categories are shown as “Reference” (italics)

Pre-COVID-19 period: 2019 survey (n = 16,253); During-COVID-19 period: 2022 survey (n = 21,444)

Statistically significant p-values (p < 0.05) are shown in bold

Health-related characteristics were strongly associated with self-medication. Bodily pain demonstrated a consistent gradient, with individuals reporting moderate or severe pain exhibiting a higher prevalence of self-medication in both periods. Long-standing illness or chronic health problems were also strongly associated with self-medication before and during the pandemic. Depressive disorder was associated with a higher prevalence of self-medication only in the pre-COVID-19 period. Limitations in daily activities were associated with a lower prevalence of self-medication before the pandemic but were not significantly associated with self-medication in the during-COVID-19 period. Body mass index showed no significant association with self-medication in the pre-COVID-19 period, while individuals with normal weight had a slightly lower prevalence of self-medication in the during-COVID-19 period compared with underweight individuals. Behavioral factors were consistently associated with self-medication. Tobacco and alcohol use were both associated with a higher prevalence of self-medication in the pre- and during-COVID-19 periods.

Patterns of health care utilization also demonstrated period-specific associations. More frequent family physician visits were associated with a higher prevalence of self-medication in the pre-COVID-19 period but showed an inverse association in the during-COVID-19 period among individuals with three to five visits. Specialist physician visits showed a modest association with self-medication only in the pre-COVID-19 period among individuals with six or more visits. Finally, unmet health care needs due to waiting lists were consistently associated with a higher prevalence of self-medication in both periods, whereas unmet need due to inability to afford health care was associated with self-medication only in the pre-COVID-19 period.

Table 3 presents the results of the pooled modified Poisson regression model with robust variance examining factors associated with self-medication across both survey waves. After adjustment for all covariates, self-medication was significantly more prevalent in the during-COVID-19 period compared with the pre-COVID-19 period (aPR = 1.11, 95% CI: 1.08–1.14). Female gender was independently associated with a higher prevalence of self-medication compared with males (aPR = 1.28, 95% CI: 1.24–1.32). Educational attainment demonstrated a clear dose–response pattern, with progressively higher prevalence ratios observed with increasing education levels: primary education (aPR = 1.19, 95% CI: 1.13–1.25), secondary education (aPR = 1.28, 95% CI: 1.21–1.36), associate/bachelor’s degree (aPR = 1.46, 95% CI: 1.38–1.55), and master’s/doctoral degree (aPR = 1.79, 95% CI: 1.65–1.95), all compared with individuals with no formal education. Age differences were modest; individuals aged 30–44 years had a slightly higher prevalence of self-medication compared with those aged ≥ 65 years (aPR = 1.06, 95% CI: 1.00–1.11), whereas other age groups did not show significant differences. Married (aPR = 1.08, 95% CI: 1.03–1.13) and divorced individuals (aPR = 1.08, 95% CI: 1.01–1.16) had a higher prevalence of self-medication compared with single individuals. Employment was associated with increased self-medication (aPR = 1.11, 95% CI: 1.04–1.18), while being outside the labor force was not significantly associated. Lack of health insurance coverage was associated with a slightly higher prevalence of self-medication (aPR = 1.06, 95% CI: 1.01–1.12). Regarding health-related characteristics, limitations in daily activities were associated with a lower prevalence of self-medication (aPR = 0.94, 95% CI: 0.91–0.97). Bodily pain showed a strong association, with higher prevalence observed among individuals reporting moderate pain (aPR = 1.21, 95% CI: 1.17–1.25) and severe pain (aPR = 1.25, 95% CI: 1.20–1.30). Individuals with long-standing illness or chronic health problems had a higher prevalence of self-medication (aPR = 1.29, 95% CI: 1.25–1.34), and depressive disorder was also associated with increased prevalence (aPR = 1.09, 95% CI: 1.04–1.15). Behavioral factors were consistently associated with self-medication, including tobacco use (aPR = 1.15, 95% CI: 1.11–1.18) and alcohol use (aPR = 1.21, 95% CI: 1.15–1.27). Body mass index was not significantly associated with self-medication. In terms of health care utilization, six or more specialist physician visits were associated with a higher prevalence of self-medication (aPR = 1.10, 95% CI: 1.02–1.19). Finally, unmet health care needs due to waiting lists (aPR = 1.15, 95% CI: 1.12–1.18) and difficulty affording health care (aPR = 1.08, 95% CI: 1.03–1.12) were both associated with a higher prevalence of self-medication.

Table 3.

Adjusted prevalence ratios for factors associated with self-medication in pooled modified Poisson regression models with robust variance

Variable		aPR (95% CI)	p
COVID-19 period	Pre-COVID-19 (2019)	reference
	During-COVID-19 (2022)	1.11 (1.08–1.14)	< 0.001
Gender	Male	reference
	Female	1.28 (1.24–1.32)	< 0.001
Educational attainment	No formal education	reference
	Primary education	1.19 (1.13–1.25)	< 0.001
	Secondary education	1.28 (1.21–1.36)	< 0.001
	Associate/Bachelor’s degree	1.46 (1.38–1.55)	< 0.001
	Master’s/Doctoral degree	1.79 (1.65–1.95)	< 0.001
Age	18–29	0.97 (0.91–1.03)	0.310
	30–44	1.06 (1.00–1.11)	0.045
	45–64	1.01 (0.96–1.06)	0.712
	≥ 65	reference
Marital status	Single	reference
	Married	1.08 (1.03–1.13)	< 0.001
	Divorced	1.08 (1.01–1.16)	0.036
	Widowed	1.04 (0.96–1.12)	0.313
Employment status	Unemployed	reference
	Employed	1.11 (1.04–1.18)	0.002
	Not in the labor force	1.01 (0.95–1.08)	0.780
Health insurance coverage	Yes	reference
	No	1.06 (1.01–1.12)	0.018
Limitation in daily activities	No	reference
	Yes	0.94 (0.91–0.97)	< 0.001
Bodily pain	None/Very mild/Mild	reference
	Moderate	1.21 (1.17–1.25)	< 0.001
	Extreme/Severe	1.25 (1.20–1.30)	< 0.001
Self-perceived health	Very good/Good	reference
	Fair	1.04 (1.00–1.07)	0.036
	Very bad/Bad	0.95 (0.90–1.01)	0.113
Long-standing illness or health problem	No	reference
	Yes	1.29 (1.25–1.34)	< 0.001
Depressive disorder	No	reference
	Yes	1.09 (1.04–1.15)	< 0.001
Body mass index (BMI)	Underweight	reference
	Normal	0.93 (0.86–1.00)	0.051
	Overweight	0.95 (0.87–1.03)	0.179
	Obese	0.96 (0.89–1.04)	0.365
Tobacco use	No	reference
	Yes	1.15 (1.11–1.18)	< 0.001
Alcohol use	No	reference
	Yes	1.21 (1.15–1.27)	< 0.001
Inpatient care utilization	0	reference
	1–2	1.04 (0.98–1.09)	0.172
	3–5	0.92 (0.84–1.00)	0.061
	≥ 6	0.95 (0.88–1.02)	0.162
Family physician visits	0	reference
	1–2	1.01 (0.98–1.04)	0.618
	3–5	0.99 (0.94–1.04)	0.746
	≥ 6	1.08 (0.99–1.18)	0.090
Specialist physician visits	0	reference
	1–2	0.99 (0.96–1.02)	0.601
	3–5	1.02 (0.97–1.07)	0.415
	≥ 6	1.10 (1.02–1.19)	0.017
Unmet need for health care due to waiting lists	No	reference
	Yes	1.15 (1.12–1.18)	< 0.001
	No need for health care	0.90 (0.80–1.01)	0.062
Unmet need for health due to inability to afford	No	reference
	Yes	1.08 (1.03–1.12)	< 0.001
	No need for health care	0.97 (0.91–1.04)	0.394

Abbreviations: aPR adjusted prevalence ratio, CI confidence interval

All covariates were included simultaneously. Reference categories are shown as “Reference” (italics)

Statistically significant p-values (p < 0.05) are shown in bold

Table 4 presents the results of the global interaction tests examining whether the associations between selected determinants and self-medication differed between the pre- and during-COVID-19 periods. Statistically significant interactions were observed for educational attainment (Wald χ² = 16.14, df = 4, p = 0.003), age group (Wald χ² = 14.46, df = 3, p = 0.002), and health insurance coverage (Wald χ² = 11.97, df = 1, p < 0.001), indicating that the relationship between these variables and self-medication varied significantly between the two periods. In contrast, no significant interaction effects were found for gender (Wald χ² = 0.83, df = 1, p = 0.363), long-standing illness or chronic health problems (Wald χ² = 1.84, df = 1, p = 0.175), or self-perceived health status (Wald χ² = 3.78, df = 2, p = 0.151). These findings suggest that while the associations of educational attainment, age, and health insurance coverage with self-medication changed between the pre- and during-pandemic periods, the effects of gender, chronic illness, and perceived health status remained relatively stable across the two survey waves.

Table 4.

Global interaction tests between COVID-19 period and selected determinants of self-medication in pooled modified Poisson regression models

Interaction term	df	Wald χ²	Global p
COVID-19 period × Gender	1	0.83	0.363
COVID-19 period × Educational attainment	4	16.14	0.003
COVID-19 period × Age group	3	14.46	0.002
COVID-19 period × Health insurance coverage	1	11.97	< 0.001
COVID-19 period × Long-standing illness or health problem	1	1.84	0.175
COVID-19 period × Self-perceived health	2	3.78	0.151

Abbreviations: df degrees of freedom

Interaction terms were tested using a global Wald chi-squared test

Statistically significant global p-values (p < 0.05) are shown in bold

Table 5 presents the stratum-specific adjusted prevalence ratios for self-medication comparing the during-COVID-19 and pre-COVID-19 survey periods among variables with statistically significant interactions in the pooled modified Poisson regression models. Stratified analyses demonstrated that the association between survey period and self-medication differed across educational attainment, age group, and health insurance coverage categories. The increase in self-medication prevalence during the COVID-19-period survey was more pronounced among individuals with an associate/bachelor’s degree (aPR = 1.18, 95% CI: 1.11–1.25), adults aged ≥ 65 years (aPR = 1.20, 95% CI: 1.11–1.29), and individuals without health insurance coverage (aPR = 1.27, 95% CI: 1.15–1.40). In contrast, comparatively smaller increases were observed among younger adults and individuals with lower educational attainment.

Table 5.

Stratum-specific adjusted prevalence ratios (aPRs) for self-medication comparing the during-COVID-19 versus pre-COVID-19 survey periods among variables with statistically significant interactions

Interaction term	Subgroup	aPR (95% CI)	p-value	Global interaction p
COVID-19 period × Educational attainment	No formal education	1.12 (1.03–1.22)	0.007	0.003
	Primary education	1.08 (1.04–1.13)	< 0.001
	Secondary education	1.10 (1.04–1.17)	0.001
	Associate/Bachelor’s degree	1.18 (1.11–1.25)	< 0.001
	Master’s/Doctoral degree	1.12 (0.97–1.28)	0.124
COVID-19 period × Age group	18–29	1.08 (1.01–1.15)	0.019	0.002
	30–44	1.05 (1.01–1.10)	0.024
	45–64	1.16 (1.10–1.21)	< 0.001
	≥ 65	1.20 (1.11–1.29)	< 0.001
COVID-19 period × Health insurance coverage	Yes	1.10 (1.07–1.13)	< 0.001	< 0.001
	No	1.27 (1.15–1.40)	< 0.001

Abbreviations: aPR adjusted prevalence ratio, CI confidence interval

Stratum-specific aPRs represent subgroup-level associations between survey period (during-COVID-19 vs. pre-COVID-19) and self-medication derived from pooled modified Poisson regression models. Global interaction p-values were obtained using Wald chi-squared tests

Statistically significant p-values (p < 0.05) are shown in bold

Discussion

Using data from national health surveys conducted in Türkiye, this study documents higher self-medication prevalence during the COVID-19-period survey and provides new evidence that the associations between self-medication and selected sociodemographic and health-related determinants differed between survey periods. Although the difference in prevalence between the pre- and during-pandemic periods appears modest in absolute terms (approximately 4% points), even such differences may represent a substantial number of individuals at the population level. Given the size of the adult population in Türkiye in 2022, this difference may correspond to approximately 2–3 million additional individuals reporting self-medication. The higher prevalence observed during the pandemic-period survey is consistent with a growing body of international research reporting greater self-medication prevalence during the COVID-19 [2, 3]. Previous studies suggest that pandemic-related disruptions to health care delivery, fear of infection in clinical settings, and uncertainty regarding access to services collectively contributed to greater self-directed management of symptoms [12, 17]. Systematic reviews conducted during the pandemic have similarly reported substantial increases in self-medication prevalence across diverse populations, with pooled estimates approaching or exceeding 45% globally, indicating that the observed rise in Türkiye reflects a broader international pattern rather than a country-specific phenomenon [2, 3, 8].

Consistent with much of the existing literature, female gender was associated with a higher prevalence of self-medication in both the pre- and during-pandemic periods [7, 9, 11]. Studies conducted in diverse settings have attributed this pattern to women’s higher health awareness, greater involvement in family health management, and more frequent use of medications and over-the-counter products [6]. Importantly, the interaction analysis did not indicate a significant modification of this association by the pandemic period, suggesting that gender-related differences in self-medication may represent a relatively stable behavioral pattern across survey periods rather than one specifically linked to the pandemic context.

Educational attainment also showed a strong and graded association with self-medication across both survey periods, with progressively higher prevalence ratios observed among individuals with increasing levels of education. Similar dose–response relationships have been documented in several European and international studies, where higher education has been linked to greater health literacy, confidence in symptom management, and autonomy in treatment decisions [8, 10, 18]. In the present study, the magnitude of this association was more pronounced during the pandemic period, and the global interaction test indicated that the relationship between educational attainment and self-medication differed significantly between survey periods. This finding suggests that educational gradients in self-medication were not entirely stable over time and may have become more marked during the pandemic period, a pattern that may reflect differences in access to health information, self-care confidence, and treatment decision-making during periods of health system disruption [6].

Age-related patterns revealed more nuanced dynamics. Adults aged 30–44 years showed higher prevalence of self-medication than older adults in both survey periods. However, the interaction analysis indicated that the strength of the association between age and self-medication differed between the pre-pandemic and pandemic periods, with relatively stronger associations observed for individuals aged 45–64 years and those aged 65 years or older during the pandemic period. Although several studies have reported lower self-medication prevalence among older adults during the pandemic [14], other research conducted during the pandemic suggests that heightened vulnerability, increased risk perception, and barriers to health care access may have influenced self-medication behaviors among older populations, particularly in contexts of prolonged health system strain [2, 12]. These findings underscore that age may function not only as a static determinant of self-medication but also as a factor whose association with self-medication can vary across different health system contexts.

Marital status was also associated with self-medication in both survey periods. Compared with single individuals, married and divorced participants showed higher prevalence of self-medication. This pattern may reflect differences in health-seeking behaviors, household health management, or the presence of medications within the household environment. Previous studies have similarly suggested that family context and shared medication practices within households may influence self-medication behaviors [19, 20].

Employment status was associated with self-medication, although the pattern differed between survey periods. In the pre-pandemic period, the association between employment and self-medication was not statistically significant, whereas during the pandemic period employed individuals showed a higher prevalence of self-medication compared with unemployed individuals. Similar patterns have been reported in several studies and are often attributed to time constraints, work-related stress, and the perceived need for rapid symptom relief among working populations [8, 14].

Health insurance coverage emerged as a determinant whose association with self-medication differed between survey periods. In the pre-pandemic period, lack of insurance was not significantly associated with self-medication, whereas during the pandemic period individuals without health insurance showed a higher prevalence of self-medication. The global interaction test further indicated that the relationship between insurance status and self-medication differed significantly between the two periods. Previous research has highlighted the role of insurance coverage in facilitating access to formal health care, particularly during periods of health system strain [18, 20]. However, some studies have reported no consistent association between insurance status and self-medication, suggesting that the influence of insurance coverage may vary across settings [2, 14]. The findings of the present study suggest that differences in access to health services during the pandemic period may have been associated with greater reliance on self-medication among individuals without health insurance.

Health-related factors were among the strongest predictors of self-medication in both periods. Bodily pain was consistently associated with a higher prevalence of self-medication, particularly among individuals reporting moderate to severe pain. Previous studies have similarly reported that pain is commonly associated with self-treatment behaviors because of its immediacy and impact on daily functioning [7, 21]. The slightly stronger associations observed during the pandemic-period survey are also consistent with studies describing disruptions in access to pain management services and greater use of over-the-counter analgesics during the COVID-19 era [3, 15].

Long-standing illness or chronic health problems were also strongly associated with self-medication in both periods. Individuals reporting a long-standing illness consistently showed a higher prevalence of self-medication compared with those without such conditions. Similar findings have been reported in studies suggesting that individuals with chronic conditions are often more accustomed to managing symptoms and medications independently, particularly when access to routine care is disrupted [2, 6, 8]. In the present study, interaction analysis did not indicate a significant modification of this relationship by the pandemic period, suggesting that the association between chronic illness and self-medication reflects a stable pattern of health-related need rather than a pandemic-specific effect.

Self-perceived health status also showed some variation across survey periods. Before the pandemic, individuals reporting fair health had a slightly higher prevalence of self-medication compared with those reporting good or very good health. However, this association was not observed during the pandemic period, and interaction analysis did not indicate a statistically significant modification of this relationship by the pandemic. Depressive disorder, on the other hand, was associated with self-medication in the pre-pandemic period but not during the pandemic period. One possible explanation is that the widespread increase in psychological distress during the COVID-19 pandemic may have reduced the ability of depressive symptoms to differentiate self-medication behaviors across individuals.

Patterns of health care utilization also differed between survey periods. Before the pandemic, more frequent visits to family physicians and specialist physicians were associated with a higher prevalence of self-medication, suggesting that individuals with greater health care utilization may also engage in concurrent self-management behaviors [8, 19]. During the later-pandemic survey period, however, these associations appeared weaker or were no longer statistically significant, and in the case of family physician visits a modest inverse association was observed among individuals reporting three to five visits. These differences may reflect changes in health care utilization patterns and access pathways across survey periods. In contrast, unmet health care needs due to waiting lists remained consistently associated with higher self-medication prevalence in both periods, supporting the potential importance of structural access barriers in self-medication behaviors [6, 8, 10, 17]. Unmet need due to inability to afford care was associated with self-medication only in the pre-pandemic period, suggesting that during the pandemic capacity-related barriers may have become relatively more salient than financial constraints in influencing self-medication behavior [7, 8].

Limitations

This study has several limitations that should be considered when interpreting the findings. First, the analyses were based on cross-sectional survey data, which precludes causal inference and limits the ability to determine temporal relationships between self-medication and its associated factors. Second, self-medication was measured using self-reported information referring to a two-week recall period, which may be subject to recall bias and social desirability bias, potentially leading to underreporting or misclassification of self-medication behaviors. Third, the outcome definition combined the use of medicines, vitamins, and dietary supplements without a prescription. These categories may represent heterogeneous behaviors with different clinical implications, and the survey does not provide information on the type of medication, dosage, indication, or appropriateness of use. Therefore, the findings should be interpreted as reflecting general patterns of non-prescribed health product use rather than inappropriate or unsafe medication use specifically. Fourth, although the THS was designed to provide national-level estimates for Türkiye, the data exclude institutionalized populations and do not capture detailed clinical information, prescription histories, medication types, or reasons for self-medication. In addition, as with all survey-based studies, the findings may be affected by survey non-response and residual confounding due to unmeasured factors that were not captured in the dataset. Furthermore, the publicly available the THS microdata did not provide sufficient survey design information required for fully design-based estimation, including primary sampling unit and stratum identifiers. Therefore, the analyses could not account for clustering, stratification, or fully specified sampling weights, and standard errors and confidence intervals should be interpreted with caution. Finally, although interaction analyses allowed assessment of whether associations between selected determinants and self-medication differed between the pre- and during-COVID-19 survey periods, the study design does not permit causal attribution of observed differences to the pandemic itself. Unmeasured contextual factors—such as changes in pharmaceutical regulation, health care access, economic conditions, media influence, or informal access to medicines between 2019 and 2022—may also have contributed to the observed patterns. In addition, multiple statistical comparisons were conducted across regression models and interaction analyses. Although the analyses were guided by predefined hypotheses and interaction testing was limited to selected variables, the possibility of chance findings due to multiple testing should be considered when interpreting isolated statistically significant associations. Despite these limitations, the use of large national health survey datasets and a comparative analysis across two survey periods strengthens the robustness and public health relevance of the findings.

Conclusion

This study provides population-based evidence on changes in self-medication in Türkiye between the pre- and during-COVID-19 survey periods. The findings indicate that the prevalence of self-medication was higher in the during-pandemic survey period and that the associations between several determinants and self-medication differed across periods. In particular, the results suggest that age, health insurance coverage, and educational attainment showed differing associations with self-medication across survey periods, whereas the associations with gender and several health-related factors remained relatively stable. These patterns may indicate that structural access to health services and underlying health needs were differently associated with self-medication behaviors across the pre- and during-COVID-19 survey periods, particularly within the later-pandemic context represented by the 2022 survey.

From a public health perspective, the findings highlight the importance of maintaining access to primary care and continuity of care for populations with greater health needs, such as older adults and individuals with chronic conditions. Ensuring timely access to health services and reducing financial or organizational barriers to care may help reduce reliance on non-prescribed medications and other self-directed health products when access to formal care becomes more limited. In addition, public health communication strategies may benefit from promoting safe medication practices and encouraging appropriate use of health services, particularly among individuals experiencing higher symptom burden or barriers to care. While the study does not allow causal attribution of observed differences to the COVID-19 pandemic, the results illustrate how different phases of a large-scale public health crisis may be associated with shifts in self-care practices and self-medication behaviors. Monitoring self-medication patterns and ensuring access to reliable health information and primary care services may therefore represent useful components of broader health system resilience strategies during periods of health system strain. Future studies should consider examining non-prescribed medicine use and the use of vitamins or dietary supplements separately, as these behaviors may have different determinants and public health implications.

Abbreviations

aPRs

Adjusted prevalence ratios

CIs

Confidence intervals

χ²

Chi-square

THS

Türkiye Health Surveys

TurkStat

Turkish Statistical Institute

WHO

World Health Organization

Authors’ contributions

The article was written by a single author.

Funding

The author declares that no funds, grants, or other support were received during the preparation of this manuscript.

Data availability

The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study was based on secondary analysis of fully anonymized publicly available survey microdata obtained from the Turkish Statistical Institute (TurkStat). According to institutional and national regulations governing secondary analysis of anonymized public-use datasets, additional ethical approval was not required.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.