Factors associated with patient adherence to medical recommendations during a health crisis
Anat Reiner-Benaim, Shimon Amar
TL;DR
This study examines how a global pandemic affects patient adherence to medical recommendations for infectious diseases, finding that social restrictions and lockdowns significantly reduce adherence to diagnostic tests and antibiotics.
Contribution
The study provides new insights into how public health crises impact patient adherence to acute infection management strategies.
Findings
Social restrictions and lockdowns were associated with decreased adherence to diagnostic tests and certain antibiotics.
Rural clinics and phone call visits were linked to lower adherence rates across all outcomes.
Adherence to prescribed antibiotics dropped for penicillin and cephalosporins but increased for macrolides during the pandemic.
Abstract
Patient non-adherence to medical recommendations is associated with significant economic and health consequences, particularly in chronic diseases. While less studied, similar implications are also observed for short-term antibiotic treatment non-adherence. A global pandemic creates a complex medical reality with the potential to impact patient adherence, and it is of interest to investigate its effect on patient adherence regarding common community-acquired infectious diseases. This study focused on primary care clinic visits with a diagnosis of fever or cough and followed the performance of referred laboratory tests and chest X-rays and the purchase of prescribed antibiotics, as adherence outcomes. A comprehensive and unique dataset of clinic visits in Southern Israel was used to assess the impact of the pandemic on patient adherence outcomes. We compared the outcomes between the…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Fig 1
Fig 2
Fig 3
Fig 4Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsAntibiotic Use and Resistance · Medication Adherence and Compliance · Vaccine Coverage and Hesitancy
Introduction
The medical literature is replete with research on adherence to medical treatment, especially in chronic diseases, and less prominently regarding antibiotic therapy [1,2] Adherence, defined as the extent to which a patient's actions coincide with the prescribed treatment regimen/advice, plays a pivotal role in achieving optimal health outcomes [3–5]. Adherence to medical treatment poses a significant challenge both for healthcare providers and patients during routine times [5]. The potential negative consequences of non-adherence such as increased health care costs, substantial worsening of disease and even increased mortality, are amplified during public health crises [6]. Public health crises introduce unprecedented stressors, including resource limitations, shifting healthcare priorities, and rapid changes in treatment guidelines, which likely impacted patient-provider communication, access to care, and the overall healthcare experience [7–9]. Furthermore, public health interventions, such as social distancing and lockdowns, while crucial for disease control, can have unintended consequences on patient routine behavior, including reduced healthcare access, increased anxiety, and disruptions to daily lifestyle habits [10].
Thus, exploring patient adherence to medical recommendations in global health crises is crucial for optimizing public health outcomes and mitigating negative impacts [11,12]. By understanding the complex interplay of the individual, the healthcare provider, and the systematic factors that influence patient behavior, we can develop effective strategies to promote adherence and improve overall health during such challenging times [13,14]. Nevertheless, while the factors influencing adherence to monitoring, balancing, and managing chronic diseases during health crises have been well studied, the specific determinants of adherence in common infectious diseases remain under-explored [12]. While research on chronic disease management during pandemics revealed disruptions to routine care, the complexities of managing common infections, including rapid disease progression, diagnostic challenges, and the need for timely interventions, require a distinct focus [15]. Due to the potential medical consequences of non-adherence or overuse of antibiotics in common infections, particularly during medical crises, understanding adherence in this context is crucial for assessing the broader health and economic implications [16].
The Coronavirus Disease 2019 (COVID-19) pandemic introduced a unique medical reality, characterized by all the above-mentioned challenges [17,18]. A particularly striking illustration of the fluctuating decision-making processes observed among both healthcare professionals and patients during the COVID-19 pandemic was the global variable uptake of COVID-19 vaccines [19–21]. This phenomenon was significantly influenced by prevailing medical uncertainties and the often-contradictory nature of media reporting on the subject. In our previous work, we delved into the responses of primary care physicians (PCPs) to these unique challenging conditions [22]. Following our findings on a decline in referral and prescription rates, the current study investigated the patient's perspective. It aimed to assess patient adherence during the COVID-19 pandemic and identify factors that affect the likelihood of adherence. We hypothesized that patients would exhibit altered adherence behaviors, including changes in their willingness to undergo diagnostic tests and their compliance with prescribed treatment regimens.
The study focused on clinic visits with a diagnosis of fever or cough, which are the two most common infectious complaints [23]. To assess patient adherence, we tracked the completion of recommended laboratory tests, chest X-rays, and the acquisition of prescribed antibiotics following primary care physician visits. We used a large dataset of visits to clinics in Southern Israel to assess the impact of the pandemic on patient adherence outcomes. We compared the outcomes between the pandemic period and the preceding and succeeding periods and used multivariate modeling to study the association of the pandemic period with adherence, subject to patient and visit characteristics.
Methods
Study population and design
Data were collected retrospectively from the database of Clalit Health Services (CHS), the largest HMO in Israel, on all visits to a PCP in Southern Israel between January 1, 2018, and December 31, 2022, that resulted in a diagnosis of either cough, fever, or both, in accordance with the International Classification of Diseases (ICD-9).
Data collection Data included patient gender, age, clinic sector, type of visit, referrals, prescriptions, performance of tests and purchase of drugs. Laboratory tests included complete blood count (CBC), c-reactive protein (CRP), and throat culture.
The laboratory tests included were the most frequent and comprised nearly 87% of all tests. Other tests were for coronavirus (6.9%), Epstein-Baar virus (EBV,5.4%), or cytomegalovirus (0.4%). Antibiotic medications included penicillin, cephalosporins and macrolides, defined in accordance with the Anatomical Therapeutic Chemical (ATC) classification.
Adherence was defined as performance of the referred laboratory tests and chest X-ray or purchase of the prescribed antibiotic medication within seven days of the visit, in which the referral or the prescription were issued. This cutoff for defining adherence was selected as it aligns with common clinical practice guidelines [24] for the course of many acute infections. Accordingly, it is expected that the patient will perform the laboratory test or purchase the prescribed medication within this time frame to achieve effective medical care.
Other measures for adherence focus mainly on treatment, including the proportion of days covered (PDC) [25,26], defined as the percentage of days within a specified period that the patient had access to the medication. However, such measures do not objectively reflect a patient's action following the physician's referral or prescription. More precise measures that are often impractical for large-scale retrospective studies include a count of pill intake [27] and self-reported adherence, which is subject to significant recall and social desirability biases [28].
The data collection period was divided into sub-periods in accordance with the social restrictions imposed in Israel during the COVID-19 pandemic, starting on March 17, 2020, and ending on May 27, 2021, which included social distancing, universal masking and three national lockdowns (Fig 1).
Study period and social restrictions timeline.Including marks of the three lockdowns imposed in Israel during the COVID-19 pandemic.
Ethical considerations
This study was approved by the CHS Research Ethics Committee with a waiver of informed consent given the use of a deidentified data source (study number 0032-21-COM2). All data were fully anonymized by CHS prior to granting access to the researchers. Data was accessed through CHS cloud platform secured by a personal password, temporarily given for the research period only. The study is reported following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies. Data was accessed for research purposes on April 9, 2023.
Statistical analysis
Patient characteristics were compared between social restriction periods using Chi-square test or Fisher's test for categorical variables, analysis of variance for normally distributed variables, and Kruskal-Wallis test for non-normally distributed variables. Normality was tested for continuous variables using the Shapiro–Wilk test. Dichotomous variables were defined for the occurrence of each adherence outcome in a visit. Trend plots marked by restriction periods were used to describe changes in adherence rates across time.
Odds ratios for the effect of each period relative to the period before the restrictions were estimated for each outcome, while adjusting for patient and visit characteristics. A multivariate generalized linear mixed effects model with the binomial family and the logit link function was fit to each adherence outcome. Random effects were used to account for repeated visits of the same patient. Gender, age, clinic sector, visit type, diagnosis, and season at visit were included in each model. We first fit main effects models to consider trends in outcomes across time with high-resolution division of time into periods, such that each lockdown was considered separately. Next, to assess trends that were specific to particular age groups and diagnoses, we fit models that included interaction effects of time period with age and diagnosis, but this time combined all lockdowns as one period, namely the “restrictions period,” to allow useful interpretability of the interaction effects. Adjusted odd-ratios were estimated for the statistically significant interactions to obtain group-specific odds ratios relative to the period before the restrictions. All adjusted interaction odds ratios were visualized on a heatmap across all outcomes and subgroups.
P-values were adjusted for multiple testing by controlling the false discovery rate (FDR) using the Benjamini-Hochberg procedure [29]. Two-sided tests with α = 0.05 were considered statistically significant. To provide a single, intuitive summary measure, we computed an overall adherence index as a visit-weighted means of the component-specific adherence indicators (CBC, CRP, throat culture, chest X-ray, penicillin, cephalosporins, macrolide). Weights equaled the share of each referral/prescription type among all opportunities within the relevant period. This metric complements (but does not replace) the outcome-specific analyses and is reported descriptively.
Results
Patients
In total, 609,823 visits to PCP by 216,819 individuals complied with the cohort definitions (Table 1). The mean age across all visits was 22.5 (±26.4) years, and 51.8% of all visits were by females (Table 1). Most visits (69.1%) took place at Jewish urban clinics, compared to 25.7% for Bedouin clinics and 5.2% for Jewish rural clinics. Telehealth visits (by phone call) were 11.5% of all visits. The period before social restrictions lasted 26.5 months, and the period after social restrictions lasted 19.5 months.
Table 1: Patient and visit characteristics.
As shown in Table 2, before and after social restrictions, 39.2%−41.2% of the visits were by children up to three years old, and 14.3%−14.9% were by adults older than 60 years old. During lockdowns, visits of young children were in lower proportions (21.9%−36.2%) and visits of older patients were in higher proportions (17.9%−20.6%). Visits by the intermediate adult age group (ages 20–59) showed a particularly large increase to 30.9%−41.4% during the social restrictions, compared to 23%−23.5% before and after social restrictions. All groups returned to their initial proportions after social restrictions ended.
Table 2: Patient and visit characteristics, compared between social restriction periods.
During lockdown periods, between 25% and 42.7% of the visits were telehealth visits. After social restrictions, such visits became more common (15%) compared to their rate before the restrictions (3%). Throughout all periods, around 4% of the visits included diagnosis of both cough and fever, with the rest of the visits divided similarly between visits with only cough diagnosis and visits with only fever diagnosis. While visits with only cough diagnosis and visits with only fever diagnosis were typically similar in proportion, which ranged between 46.2% and 49.2%, the first and third lockdowns were characterized by a shift towards more visits with only cough diagnosis (56.3%−57.8%), while the second lockdown was characterized with a shift towards visits with only fever diagnosis (52.2%).
Adherence outcomes
As shown in Table 1, 8.4% of the visits resulted in a referral to a chest X-ray, 16.2% of the visits resulted in a referral to laboratory tests, and 8.1% of the visits resulted in prescribing antibiotic drugs. Complete blood count was the most frequently referred laboratory test (12.9%) and penicillin was the most frequently prescribed antibiotic drug (5.1%). We studied the associations of these referrals and prescriptions with the pandemic period in our previous paper [22].
Changes in patient adherence rates during the pandemic period varied between outcomes (Table 2 and Fig 2). Adherence to laboratory test referrals showed sporadic drops once social restrictions started. The performance rate dropped from 59.1% to 53.1% for CBC, from 67.9% to 60.8% for CRP, and from 92.2% to 81.5% for throat culture. Nevertheless, the performance rate of chest X-ray did not decrease during the social restrictions, remaining between 65.7% and 66.6%, yet it dropped during the period after social restrictions to a level as low as 50% (Fig 2).
Patient adherence rates (%) across time (monthly).Includes marks of beginning and end of social restrictions period (dotted green lines) and the three lockdown periods (grey). Adherence to laboratory test referrals showed sporadic drops once social restrictions started. Adherence dropped for CBC, CRP, and throat culture. Adherence to penicillin and cephalosporins dropped during the second lockdown but increased after social restrictions. Adherence to macrolides increased during the first lockdown but dropped during the next lockdowns.
In contrast to the referred tests, adherence to antibiotic prescriptions was more substantially changed during the pandemic period. The Purchase rate of penicillin showed a major drop from 30.5% before the social restrictions to as low as 9.7% during the second lockdown, and an increase to as high as 50%, well beyond the initial levels, after social restrictions (Fig 2). A similar trend was observed for cephalosporins, which dropped from 82% adherence before the restrictions to 64.1% during the second lockdown, and increased to 82.4%, nearly the original level, after social restrictions.
A different trend was observed for macrolides, as their purchase rate increased to 57.7% during the first lockdown, compared to 44.9% before the restrictions. However, during the next lockdowns adherence dropped to near the original levels (41.2%−44.2%). Interestingly, the decline continued after the social restrictions ended, reaching a new nadir point of 10% (Fig 2).
Overall adherence was calculated as a weighted average of adherence rates across all outcomes, with weights proportional to the number of referrals or prescriptions issued for each outcome. While before social restrictions, the overall adherence rate was 45%, during social restrictions it declined to 40% during the first lockdown, dropped to 21% during the second lockdown, and then increased to 32%. After social restrictions ended, it rose to a peak of 51%.
Factors affecting adherence during and after the pandemic
The adjusted odds ratios provided precise measures for trends in patient adherence, in comparison to the period before the restrictions, while accounting for all patient and visit characteristics. Full models result, including odds ratios calculated for interactions with age and diagnosis, are provided within S1 and S2 Tables.
Odds ratios by subgroups are presented in Fig 3 and corresponding age-specific trends in adherence are shown by Fig 4. For chest X-ray, adherence decreased during social restrictions for patients of age 60 or older (OR=0.70, S2 Table and Fig 3B) and further decreased after social restrictions (OR=0.71, Table 3 and Fig 3A), most prominently for patients of ages 0–3 (Figs 3B and 4) with only cough diagnosis (Fig 3C). For CBC test, adherence decreased during the first lockdown (OR=0.79, Table 3 and Fig 3A), most prominently for patients of ages 0–3 (Figs 3B and 4) with only cough diagnosis (Fig 3C) and returned to previous levels afterwards. For throat culture, adherence decreased during the third lockdown and onward (OR=0.64, Table 3 and Fig 3A).
Table 3: Adjusted odds ratios for time-period effect on patient adherencea.
Heatmap of adjusted odds ratios (aOR) with 95% confidence intervals (CI) for patient adherence, relative to the period before social restrictions.Adjusted odds ratios are presented on a log scale, centered at zero. Red indicates a decrease, while green indicates an increase. aOR were estimated by multivariate generalized linear mixed effects models with the binomial family and logit link. A. aOR during each lockdown and after social restrictions, relative to the period before the restrictions, adjusted for gender, age, clinic sector, visit type, diagnosis, and season at visit (see also Table 3). B-C. aOR during and after social restrictions, stratified by age (B) and by diagnosis (C), calculated based on the interaction effect with period, with adjustment for all other covariates. D. aOR for rural and Bedouin clinics, relative to urban clinics (left panel) and for telehealth visits (by phone call), relative to in-person visits. Full models’ results are provided in Supplementary S1 and S2 Tables, including regression coefficients, aOR and 95% CI.
Patient adherence rates (%) across time (monthly) for the youngest and oldest age groups.For chest X-ray, adherence decreased during social restrictions for patients aged 60 years or older and decreased after social restrictions for patients aged 0-3 years. For CBC referrals, adherence decreased during the first lockdown for patients of ages 0-3. For penicillin prescription, adherence decreased during social restrictions and increased afterwards for children aged 0-3. Opposite trends were observed for macrolides.
FDR – False discovery rate
CBC – Complete blood count
The odds ratios for patient adherence to purchasing prescribed antibiotic treatments further clarified the opposite trends between the different types of drugs. Adherence decreased for penicillin during the restrictions (OR of 0.26–0.44, Table 3 and Fig 3A), most prominently for children of ages 0–3 (Figs 3B and 4) for only cough diagnosis (Fig 3C). It also decreased for cephalosporins during the second lockdown (OR=0.52, Table 3 and Fig 3A). In contrast, purchase of macrolides increased substantially during the first lockdown (OR=2.16, Table 3 and Fig 3A), except for patients of ages 60 or older (Figs 3B and 4).
After social restrictions, the opposite trends for penicillin and macrolides were further observed, as adherence for penicillin increased substantially (OR=2.35, Table 3 and Fig 3A), particularly for children (Figs 3B and 4) with cough or fever (Fig 3C), while adherence to macrolides decreased (OR=0.33, Table 3 and Fig 3A), particularly among children of age 0–3 (Fig 3B and 4) with cough only (Fig 3C).
Clinic sector (Jewish urban, Jewish rural or Bedouin) and visit type (regular or by phone call) were found to impact patient adherence (Table 3 and Fig 3D). Visits to rural clinics were associated with lower adherence in most outcomes (OR of 0.67–0.94, Fig 3D), and visits to Bedouin clinics were associated with higher adherence in all test outcomes (Fig 3D, left panel), particularly throat culture (OR=2.53), and lower adherence in drug purchase (OR of 0.66–0.76). Telehealth visits (by phone call) were associated with lower adherence to most outcomes (Fig 3D, right panel), most prominently throat culture (OR=0.32) and cephalosporins (OR=0.45), but also chest X-ray (OR=0.85), CBC (OR=0.92), CRP (OR=0.80) and penicillin (OR=0.85).
Discussion
This large-scale retrospective cohort study investigated patient adherence, as defined by response to diagnostic tests referrals and to antibiotic prescriptions. It compared adherence patterns prior to the COVID-19 pandemic period with those observed during and after social restrictions and lockdowns.
Analyzing patient adherence to ancillary tests and antibiotic treatments during the 26.5 months preceding the COVID-19 pandemic revealed varying adherence rates. Adherence to throat cultures was highest, followed by chest X-rays, then CRP levels, and finally, CBC. A similar hierarchy was observed for antibiotic treatments, with cephalosporins exhibiting higher adherence rates than macrolides and penicillin.
Continuous analysis of adherence during the following 33.5 months, encompassing the COVID-19 pandemic and the period succeeding it, revealed decreased adherence during social restrictions. This decline was particularly notable for chest X-rays among patients older than 60, CBC among children aged 0–3, and throat cultures across all age groups. Adherence to penicillin regimens, especially among children aged 0–3, also decreased during restrictions, while cephalosporins adherence initially decreased and subsequently rebounded, reaching new peaks after restrictions were ended. Conversely, macrolide prescription adherence increased during restrictions and decreased afterward. Lower adherence rates were associated with visits to rural and Bedouin clinics, as well as phone consultations.
The COVID-19 pandemic significantly impacted medication adherence and healthcare services for patients with chronic conditions. Studies examining non-adherence among patients with chronic non-communicable diseases during the pandemic identified psychological, social, and environmental factors [6]. Impairments in mobility, freedom of movement, social connections, limited access to healthcare facilities, fear of infection, and disruptions in medication supply chains significantly impacted the ability to maintain a balance in those diseases, thereby requiring interventions across multiple medical and behavioral domains [13]. In Ethiopia, a study found that 72% of diabetic and hypertensive patients had poor medication adherence during the pandemic, with 57% reporting negative impacts on follow-up visits, medication availability, or affordability [30]. Conversely, a Korean study showed improved medication adherence and persistence among patients with dyslipidemia during the pandemic, with an increased proportion of days covered and lower risks of medication discontinuation [31].
Patient adherence is a multifaceted behavior influenced by a complex interplay of factors, both in routine and emergency settings. Under normal circumstances, non-adherence is associated with determinants such as forgetfulness, accessibility issues, concerns about side effects, perceived treatment inefficacy, low health literacy, stigmatization, lifestyle choices, financial constraints, psychological factors, inadequate social support, and suboptimal physician-patient communication [32] However, during public health emergencies like the COVID-19 pandemic, these challenges are exacerbated and compounded by additional stressors. Disruptions in medication supply chains, economic hardships, restricted access to healthcare services, heightened fear and anxiety, and communication barriers become prominent factors affecting adherence. Therefore, interventions aimed at improving patient adherence must consider both the chronic determinants and the acute challenges posed by emergency situations, ensuring tailored strategies that address the unique needs of individuals in diverse contexts [12,33].
The aforementioned factors may help elucidate our findings, starting with those observed before the pandemic. Patients’ perception of illness severity, coupled with their confidence in the diagnosis and treatment, led to higher adherence rates for broad-spectrum antibiotics, such as cephalosporins and macrolides, that are often prescribed for more severe infections, compared to penicillin [34,35]. Considerations of convenience and accessibility explain the higher compliance rates for throat cultures, performed in the doctor's office, compared to chest X-rays and blood tests, which typically involve less availability, more effort on the patient's part, or discomfort, such as in blood tests [36]. Conversely, when symptoms are more significant, such as fever and cough, and medical evaluation is better communicated, there will be higher adherence to diagnostic tests and treatment [37].
The global medical uncertainty that predominated during the early stages of the COVID-19 pandemic led in part to empirical decision-making, both in diagnosis and treatment. Common symptoms of typical infectious diseases also aligned with those of COVID-19. In our previous work [22], we discussed the impact of COVID-19 on physicians’ decisions, decreasing both prescribed antibiotics and ancillary tests [38,39]. The pandemic exacerbated existing barriers to healthcare access, particularly for vulnerable populations. Factors such as convenience, support systems and perceived risk influenced patients’ decisions to seek care, leading to decreased adherence to medical recommendations, especially among the elderly and young children. In addition, patients were influenced by media reports. For instance, macrolides were briefly perceived as effective in inhibiting inflammatory processes in COVID-19, which may explain the increased patient adherence to this treatment [40,41]. Adherence to throat culture and penicillin and cephalosporins treatments was deemed unnecessary due to the increasing incidence of COVID-19 and the fact that sore throat was one of its symptoms.
Following the lockdowns and restrictions, our initial study [23], observed changes in infectious disease patterns, with a corresponding rise in common illnesses such as pharyngitis. Indeed, in the current study we found a significant increase in patient adherence to cephalosporins and penicillin treatment during the post-restrictions period and return to nearly pre-restrictions levels in adherence to throat culture. This further highlights the interpretation that in the absence of active epidemic and conflicting medical information being disseminated to the public, regardless of accuracy, the public is more likely to heed medical recommendations, even for relatively minor infectious diseases.
Furthermore, we observed “delayed effects” of treatment adherence during the post-restrictions period. For instance, adherence to macrolides decreased significantly, possibly due to lingering negative perceptions of the treatment arising from the pandemic period, when conflicting messages were circulated regarding its efficacy versus its contribution to antibiotic resistance. Similar behavior was observed in the public's perception of chest X-rays being ineffective in diagnosing the cause of a cough, as was the case during the pandemic. We can venture to speculate that these negative memories will fade over time, and future adherence will return to pre-pandemic levels.
Despite the active patient involvement in decision-making, we believe it does not reflect a lack of trust in primary care but results from a period of heightened uncertainty causing low health literacy, along suboptimal physician-patient communication [42]. This claim is supported by our finding of difference in adherence between in-person and telemedicine consultations, and of consultation rates and adherence to most tests and treatments returning to their original levels in the post-pandemic period. We can speculate that the media played a significant role in shaping public perception of the pandemic, which in turn influenced patient behavior and treatment decisions, as well as subconsciously during physician-patient interactions in a period of heightened uncertainty. Patients made informed decisions based on the information they had accumulated at that time. Fortunately, at the end of the period, no increases in complications or mortality were observed in the study population, which may question the real need of some regimens given in first place, but not future resistant microbial strains [38,39,43]. If this is the case, we can positively leverage the influence of the media on patient decision-making and treatment adherence in the future [44].
Nevertheless, the media's dissemination of unsubstantiated information during times of crisis influenced public perception of effective treatments [45]. Amidst the scientific and medical uncertainty surrounding the pandemic, the public was bombarded with claims regarding the efficacy of various interventions, including macrolides, vitamins, steroids, and turmeric. Despite the lack of rigorous scientific evidence, these claims gained traction, particularly among those experiencing fear and helplessness. Consequently, it proved challenging to dissuade the public from pursuing these unproven treatments, which often relied on the placebo effect. The absence of readily available, official, and expert-vetted information during crises creates a void that is naturally filled by rumors and unverified claims. This gap between the public's need for knowledge and the lack of credible sources exacerbates the problem of misinformation. When individuals are anxious and uncertain, they are more susceptible to believing in unsubstantiated claims, particularly when these claims offer hope or a sense of control [17,46].
Our study revealed marked deviations in antibiotic adherence within the youngest (0–3 years) and oldest patient cohorts during the COVID-19 pandemic, reflecting a confluence of age-specific vulnerabilities. For young children, adherence is inherently dependent on caregiver actions, which were likely influenced by heightened pandemic-related anxieties, disruptions to routine care, and potential hesitancy toward medical interventions amid widespread misinformation [47]. Caregivers may have faced an increased burden due to childcare closures and economic strain, impacting their ability to adhere to medication regimens. In contrast, older adults, who often manage multiple comorbidities, were likely to encounter barriers related to reduced mobility, fear of infection at healthcare facilities, and potential cognitive decline, further compounded by social isolation during lockdowns [48]. Moreover, both groups are susceptible to communication challenges, with young children relying on caregivers’ interpretation of medical advice and older adults potentially experiencing difficulties in understanding or remembering instructions. These findings align with broader research indicating that pandemic-induced disruptions disproportionately affected vulnerable populations, leading to reduced healthcare utilization and adherence to prescribed treatments [49].
As noted, visits to rural and Bedouin clinics, as well as phone-call consultations, were associated with lower adherence rates. During pandemics, when uncertainty is high and conventional medical treatments may seem inadequate, some rural and Bedouin populations may exhibit a heightened reliance on traditional healing practices. This trend is particularly pronounced in conservative communities where belief in traditional knowledge is deeply rooted. Limited access to healthcare due to long distances, transportation challenges, inflexible clinic hours, and socioeconomic factors such as poverty and low health literacy can also complicate adherence. While telehealth offers potential benefits, its implementation in these communities presents new challenges, including communication barriers and language disparities, potentially further distancing these populations from conventional healthcare [50]. Consequently, trust in official information erodes, adherence to guidelines diminishes, and existing health disparities widen. It is crucial to note that these phenomena are not uniform across all rural and Bedouin communities. Significant variations exist among different groups, influenced by their unique sociocultural and economic characteristics, especially during times of crisis. Understanding the cultural, psychological, and socioeconomic dimensions of this phenomenon is essential for developing effective intervention strategies [51,52].
A proportion of the non-adherence observed during restrictions likely reflects contextual constraints (financial, geographic, organizational, and information-related) rather than individual preferences alone. We therefore interpret sector- and visit-type differences as indicators of structural barriers to access and communication during crisis conditions, not solely as behavioral choices.
Strengths and limitations
Our study has several notable strengths. First, it leverages a large and diverse cohort of more than 600,000 primary-care visits across pre-pandemic, pandemic, and post-pandemic periods, providing high statistical power and representation of vulnerable subgroups such as young children, older adults, rural residents, and Bedouin populations. Second, the use of comprehensive electronic health records from a single health system ensured consistent capture of referrals, diagnostic test performance, and antibiotic purchases, thereby minimizing information bias [22,23]. Third, the application of multivariable mixed-effects models with interaction terms enabled evaluation of heterogeneous effects across patient subgroups while accounting for repeated visits [29].
At the same time, several limitations should be acknowledged. First, as an observational study, our analyses cannot establish causal relationships and should be interpreted as associations. Second, we lacked contextual variables such as comorbidities, socioeconomic status, or access-to-care indicators, leaving potential residual confounding [6,12,13]. Third, adherence was approximated by completion of tests and medication purchase, which may not fully reflect ingestion or therapeutic compliance [53]. Fourth, some care episodes may have occurred outside the primary-care setting, though the requirement for physician-issued antibiotic prescriptions in Israel reduces the likelihood of missing large numbers of antibiotic purchases. Finally, while our findings are robust within the Israeli healthcare system, their generalizability to other healthcare systems remains to be established [30,31,33].
In conclusion, this study offers a unique contribution to our understanding of patient adherence to acute infection management strategies during a pandemic, specifically COVID-19. Examining a large and diverse population over five years, we identified significant variations in adherence to both diagnostic testing and antibiotic treatment, influenced by the evolving pandemic landscape, prevailing disease patterns, and associated medical uncertainty [54].
These findings have critical implications for both antibiotic stewardship and pandemic preparedness [26,38,55]. Our results underscore the need for tailored interventions to address the specific challenges faced by diverse populations, including children, older adults, and cultural groups such as Bedouins and residents of rural communities. Such interventions may include stricter medical criteria for prescribing and dispensing certain antibiotics during pandemics, implemented for limited durations and based on accumulating evidence. Critically, effective communication is essential, encompassing both national-level public health messaging and clear, consistent communication between healthcare teams and their patients. Well-informed and confident medical professionals can effectively convey scientific evidence to patients, who are generally receptive to their guidance. Robust communication strategies are therefore crucial for addressing patient concerns, combating misinformation, and improving adherence during periods of heightened anxiety [56].
Furthermore, this study highlights the potential for digital health interventions, such as mobile health applications, to support adherence tracking and personalized care during public health crises [54]. Future research should also explore adherence among vulnerable populations, including immuno-compromised patients, individuals with disabilities, and those with chronic complicated diseases. Comparative studies with other countries are necessary to assess the generalizability of these findings and inform global pandemic preparedness and response strategies.
In periods of increasing uncertainty, more effort is warranted to promote better adherence to therapy, especially when patients try to self-manage their disease. Our results further highlight the relationship between patient preference and adherence, the complex nature of adherence and the need for adequate patient education. In light of our findings, actionable recommendations for improving adherence in future crises [57] may include implementing patient reminders for test completions, enhancing telehealth platforms for better communication, and adjusting public health campaigns to address specific misconceptions and sub-populations.
Supporting information
S1 TableMain effects models.An excel table that includes regression coefficients and odds ratios for main effects models, with 95% confidence limits and FDR-adjusted p-values, for each adherence outcome.(CSV)
S2 TableMain effects and interaction effects models.An Excel table that includes regression coefficients and odds ratios for main effects and interaction effects models, with 95% confidence limits and FDR-adjusted p-values, for each adherence outcome.(CSV)
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Losi S, Berra CCF, Fornengo R, Pitocco D, Biricolti G, Federici MO. The role of patient preferences in adherence to treatment in chronic disease: a narrative review. Drug Target Insights. 2021;15:13–20. doi: 10.33393/dti.2021.2342 34785884 PMC 8591552 · doi ↗ · pubmed ↗
- 2West LM, Cordina M. Educational intervention to enhance adherence to short-term use of antibiotics. Res Social Adm Pharm. 2019;15(2):193–201. doi: 10.1016/j.sapharm.2018.04.011 29685459 · doi ↗ · pubmed ↗
- 3Sluijs E, Dulmen S, Dijk L, Ridder D, Heerdink RR, Bensing J. Patient adherence to medical treatment: a meta review. 2006.10.1186/1472-6963-7-55PMC 195582917439645 · doi ↗ · pubmed ↗
- 4Bissonnette JM. Adherence: a concept analysis. J Adv Nurs. 2008;63(6):634–43. doi: 10.1111/j.1365-2648.2008.04745.x 18808585 · doi ↗ · pubmed ↗
- 5Burnier M. The role of adherence in patients with chronic diseases. Eur J Intern Med. 2024;119:1–5. doi: 10.1016/j.ejim.2023.07.008 37479633 · doi ↗ · pubmed ↗
- 6Hassan S-U-N, Zahra A, Parveen N, Khatoon F, Bangi NA, Hosseinzadeh H. Quality of Life and Adherence to Healthcare Services During the COVID-19 Pandemic: A Cross-Sectional Analysis. Patient Prefer Adherence. 2022;16:2533–42. doi: 10.2147/PPA.S 378245 36147381 PMC 9488595 · doi ↗ · pubmed ↗
- 7Souvatzi E, Katsikidou M, Arvaniti A, Plakias S, Tsiakiri A, Samakouri M. Trust in Healthcare, Medical Mistrust, and Health Outcomes in Times of Health Crisis: A Narrative Review. Societies. 2024;14(12):269. doi: 10.3390/soc 14120269 · doi ↗
- 8Filip R, Gheorghita Puscaselu R, Anchidin-Norocel L, Dimian M, Savage WK. Global Challenges to Public Health Care Systems during the COVID-19 Pandemic: A Review of Pandemic Measures and Problems. J Pers Med. 2022;12(8):1295. doi: 10.3390/jpm 12081295 36013244 PMC 9409667 · doi ↗ · pubmed ↗