Pediatric Medication Noninitiation in Spain

The Strengthening the Reporting of Observational Studies in Epidemiology statement and the European Symposium for Patient Adherence, Compliance, and Persistence Medication Adherence Reporting Guideline were followed in the reporting of this research.^27,28 

This was an observational study of a cohort of pediatric patients receiving at least 1 new prescription (July 2017 to June 2018) in Catalonia (Spain). Data (July 2016 to December 2018) were obtained from the Public Data Analysis for Health Research and Innovation Program real world database,²⁹  which has data from all providers in the public health system, including information on the use of health care resources, clinical information, and medication prescription and dispensing.

The Catalan public health system provides universal health care to all residents (∼7.5 million) and is organized into health areas (smaller territories that manage health care provision). In Supplemental Information 1, we give a detailed description of the characteristics of the health care system. All citizens have a unique individual identification number, which grants access to the whole public health system in Catalonia and Spain. It is funded through taxes and free of charge at the point of use, except for prescription medications, which are subjected to a copayment contribution ranging from 0% to 60% according to the type of medication, level of income, and status as a pensioner.³⁰  An electronic prescription system registers all electronic prescriptions and dispensing of publicly financed medications, which are dispensed in community pharmacies by using the Anatomical Therapeutic Chemical (ATC) classification system. Patients can get prescriptions from private providers, but these are not funded by the Catalan public health system nor registered in the databases. Primary care (PC) is the gateway to the system and provides access to secondary care (SC).

Data were anonymized, and no informed consent was needed from participants. The Fundació Sant Joan de Déu Ethics Committee approved the study (PIC-118-18).

The study included patients (<18 years old) who were prescribed a new pharmacologic treatment. The most prescribed and clinically relevant pharmaceutical subgroups were included (Table 1). The prescription was considered new when there were no prescription records for the same pharmaceutical subgroup in the previous 6 months (preperiod). Consequently, the same patient could be included a maximum of 2 times for the same pharmaceutical subgroup. No other inclusion criteria were applied.

For time-dependent variables, information recorded at the time of prescription was used.

Initiation is a time-to-event variable with a well-defined time origin (prescription) and an easily identifiable end point (dispensing).⁴  A prescription was considered noninitiated if there was no dispensing record during the follow-up period (1 month). As a sensitivity analysis, the follow-up period was extended to 6 months.

In Catalonia, patients are considered pediatric until they reach the age of 15 years in PC and 18 years in SC. The study population was stratified into age categories according to International Council for Harmonization guidance,³¹  although the adolescents age group was divided into 2 categories: 0 to 1 years old; 2 to 4 years old; 5 to 11 years old; 12 to 14 years old; 15 to 17 years old.

Patients were classified on the basis of their copayment level (assigned to the parent or legal guardian) for the year of prescription as defined by the Spanish Government³² : 0% (pensioner and nonpensioner: annual income up to ∼€5000); 10% (pensioner: annual income ∼€5000–100 000); 40% (nonpensioner: annual income ∼€5000–18 000); 50% (nonpensioner: annual income €18 000–100 000); 60% (pensioner and nonpensioner: annual income >€100 000). There is a monthly income-based ceiling for pensioners (ie, when a pensioner reaches the ceiling, the cost of subsequent dispensed medications is €0). Some drugs (such as chronic treatments) have reduced contribution (10% copayment capped at €4.26 per prescription).³²  Patients’ copayment level was estimated on the basis of their contribution to the cost of the medication.

The proportion of medication costs assumed by the patient was categorized on the basis of the distribution of the variable (Supplemental Information 2; €0, >€0–2, or >€2). The public health system (partially or totally) covers the cost of medications prescribed to patients and implements policies to reduce the cost of medication (Supplemental Information 1). Patients with different copayment levels may contribute the same (eg, patients from different copayment groups purchasing reduced contribution medications or patients who reached the monthly ceiling).

Active diagnoses at the time of prescription were considered on the basis of the International Classification of Diseases, 10th Revision, (ICD-10). The ICD-10 chapter code was used to define the category of all diagnoses, except for diabetes mellitus, behavioral and emotional disorders with onset in childhood, and asthma (Table 2) that were considered separately (on the basis of their exact coding) because of their clinical relevance.

Other variables included were patient’s sex; appointments with the PC social worker (during the study period [July 2017 to June 2018]); number of new prescriptions, visits to a PC clinician, and visits to a PC nurse (the latter 3 variables refer to the year before the new prescription assessed); specialty of the prescribing clinician (pediatrician [including PC and SC], PC clinician [includes general practitioners and family physicians], or SC clinician); characteristics of the center (PC center, after-hours PC center, or SC center); and health area.

Analyses were conducted by using Stata/MP 13.1 (Stata Corp, College Station, TX).

The unit of analysis was the prescription. The prevalence of overall noninitiation was expressed as the proportion of new prescriptions not filled within 30 days (1-month noninitiation) and 180 days (6-month noninitiation) of the date of the prescription. The prevalence of noninitiation by pharmaceutical subgroup and age group was also estimated. When the number of new prescriptions was <50, prevalence in that age group was not estimated.

To identify noninitiation explanatory factors, all available variables were included in a mixed-effects logistic regression model in which level 1 was the prescription and level 2 was the health area, by using the “melogit” command, which fits mixed-effects models for binary and binomial responses. In this model, prescriptions were clustered within health areas. One-month noninitiation was the dependent variable. With the exception of medication cost, for which the reference category was €0 to ease interpretation, for categorical independent variables, the category with the highest sample size was used as the reference group. The strength and direction of the association were reported by using adjusted odds ratios (ORs) and 95% confidence intervals (CIs). Because of the large sample size, most explanatory factors are statistically significant at 95%. A ssociations were considered clinically relevant when they were statistically significant (P < .05) and showed high effect size in categorical variables (OR: <0.9 or >1.1) or in continuous variables (OR: <0.99 or >1.01).

When no dispensing records or only reduced contribution drugs (fixed at 10%) were dispensed the year when the prescription was issued, copayment level was missing (4.16% patients). To deal with this, when available, we assigned the level of copayment from the previous or subsequent year (1.66% of cases). When this was not possible, we imputed the median copayment level of the patients’ health area (2.5% of cases). Furthermore, 18.71% of prescriptions had no active diagnosis related to the prescription recorded at the moment of prescription (Supplemental Information 2). These data were considered to be missing at random. An imputed database was generated by using multivariate imputation with chained equations by using all the available variables in the model.

A total of 1 539 003 new prescriptions were issued to 715 895 children, consisting of almost equal numbers of boys and girls aged 0 to 1 years (8.7%), 2 to 4 years (19.8%), 5 to 11 years (38.8%), 12 to 14 years (14.6%), and 15 to 17 years (18.1%). Tables 2 through 4 detail the characteristics of the sample.

Table 1 presents the prevalence of noninitiation. The overall prevalence of 1-month noninitiation was 9.0% whereas the prevalence of 6-month noninitiation was slightly lower (8.5%).

By pharmaceutical subgroups (Table 1), the highest 1-month noninitiation prevalence was observed in proton pump inhibitors (21.5%) and benzodiazepine derivatives (19.4%), whereas the lowest was observed in oral antibiotics (2.6% in penicillins with extended spectrum; 3.9% in macrolides) and 3.9% in oral glucocorticoids.

(Supplemental Table 6 (Supplemental Information 2) shows 1-month noninitiation prevalence by age groups.

Table 5 shows the noninitiation explanatory factors based on data gathered from electronic health records.

Children aged 5 to 11 years showed the lowest noninitiation risk, whereas 15- to 17-year-olds showed the highest (OR: 1.50; 95% CI 1.48–1.53).

Patients who had a 10% (OR: 0.45; 95% CI 0.44–0.47), 50% (OR: 0.76; 95% CI 0.75–0.77), or 60% copayment level (OR: 0.73; 95% CI 0.67–0.80) showed lower noninitiation odds than patients copaying 40%, whereas those with a 0% copayment level (OR: 5.18; 95% CI 4.16–6.45) showed higher odds.

Prescriptions that were free of charge were less likely to be noninitiated, and patients who received a higher number of new prescriptions during the year before the index prescription had a lower probability of noninitiation (OR: 0.92; 95% CI 0.91–0.92).

Children suffering from diseases of the respiratory system excluding asthma (OR: 0.75; 95% CI 0.75–0.76) and diseases of the eye and adnexa or of the ear and mastoid process (OR: 0.87; 95% CI 0.86–0.88), had a lower probability of noninitiation, whereas those suffering from conditions because of external causes had a higher probability of noninitiation (OR: 1.22; 95% CI 1.19–1.24).

Prescriptions issued in a SC (OR: 1.53; 95% CI 1.50–1.55) were less likely to be initiated than those issued in a PC center, whereas prescriptions made by a pediatrician were more likely to be initiated than ones issued by a PC clinician (OR: 1.16; 95% CI 1.15–1.18) or a SC clinician (OR: 1.55; 95% CI 1.51–1.59).

Our study involved an extensive overall analysis of pediatric noninitiation in several medication groups, showing different prevalence between them. The prevalence of noninitiation of antiinflammatory medications was in line with that of previous studies,¹²  whereas the prevalence of antidepressant, antimicrobial, and antiasthmatic noninitiation in other studies was almost 5 times higher.^{12,18–20,33}  Differences in setting (eg, organization of the health care system and low out-of-pocket medication cost for patients in Catalonia) or study methodologies (eg, preperiod, sample size, and study length) could explain these figures. Poor definition of the parameters that define adherence (initiation, implementation, or persistence) leads to heterogeneous adherence figures that are open to interpretation.²⁶ 

Pediatric noninitiation in the current study showed lower rates than in adults for most medication groups.¹⁰  Parents’ greater concern for their children’s health may explain lower noninitiation rates in this population^34,35  and, additionally, some medicines for children (such as syrups) have shorter expiration dates, preventing accumulation of surplus medications.³⁶  Finally, children’s medicines are sometimes prescribed pro re nata, which may partially explain the high rates of noninitiation of benzodiazepine derivatives in children up to 5 years old, commonly used to treat feverish convulsions.³⁷ 

In line with studies in adults,^8,11  overall noninitiation barely decreased when the follow-up period was extended to 6 months, and this may be because of greater representation of medication for acute conditions. Noninitiation rates of chronic treatments, such as antipsychotics, psychoanaleptics, and antidiabetics, noticeably diminished when the follow-up period was extended (sensitivity analysis), indicating a period when caregivers and patients consider acceptance of the medication. In psychoanaleptics and antipsychotics, this may be related to stigma and caregiver preference for nonpharmacologic interventions.^38,39 

Fear of side effects, preference for nonpharmacologic interventions, and stigma may also explain the decrease in 6-month noninitiation of oral antidiabetics, which are indicated in the pediatric population with overweight associated with hyperandrogenemia and polymicrocystic ovary syndrome as an off-label drug to treat early puberty in 5- to 11-year-old girls.^40,41  In insulins, however, the decrease in noninitiation over time may be because clinicians usually provide the first unit of insulin pens. Therefore, 6-month noninitiation rates may be a more reliable prevalence measure for chronic pharmaceutical treatments.

A curvilinear relationship between age and noninitiation was observed with the higher rates of noninitiation in the youngest and oldest population groups. In the youngest, this could be influenced by the fear of exposing infants to the occasional toxic effects of medications and/or by a preference for complementary alternative medicines, which are perceived as safer.⁴²  Higher rates of noninitiation prevalence in adolescents, who showed a similar rate of noninitiation to adults,¹⁰  may be explained by a lower perception of disease severity threat, their own beliefs about the need for medication, and stigma.⁴³ 

The medication cost share was the variable with the greatest impact on noninitiation, and cost of treatment has consistently been reported as a factor that can lead to nonadherence.^26,32  In line with previous studies,³²  our study revealed that even small copayments were associated with increased odds of noninitiation. Patients who did not pay for their prescriptions could be those who were exempt from contributions or those who already reached their monthly cost ceiling. In our data, 86.7% of free-of-charge prescriptions were issued to patients in the 0% copayment group and 13.3% were issued to other copayment groups. Having controlled for the cost share, patients exempt from copayment showed the highest probability of noninitiation. In other words, although these patients do not have to contribute to the treatment cost, they have the highest risk of noninitiation. To facilitate interpretation of these results, supplementary regression models were fitted (Supplemental Tables 10 and 11). When the effect of copayment on noninitiation was explored in a bivariate analysis, patients exempt from copayment showed the lowest odds of noninitiation. When this association was adjusted for medication cost assumed by the patient, the odds of noninitiation attributed to the medication cost increased and patients exempt from copayment showed the highest odds of noninitiation. We hypothesize that this occurred because the protective effect of payment exemption in this profile of patients was controlled. Patients with 0% copayment are patients with a low socioeconomic status, which may be associated with a lower educational level and higher risk of social isolation. The association found between appointments with a social worker and the risk of noninitiation might support previous arguments. In Sweden, noninitiation was associated with socioeconomic disadvantages, in the same way as lack of trust in health care and a long-term illness.⁴⁴ 

Prescriptions issued by pediatricians were more likely to be initiated than those issued by PC clinicians and SC clinicians, which may be explained by the existing bond of trust with the prescribing professional.^39,44,45 

As far as we know, this is the largest study to assess the prevalence of noninitiation and its determinants in the pediatric population and the first study to analyze noninitiation by not yet studied pharmaceutical subgroups, age groups, and multiple follow-up periods. It is also the first to identify multiple noninitiation explanatory factors related to age, copayment level, cost Programa de analítica de datos para la assumed by the patient, and clinician specialty.

The main strengths of this study are its representativeness and sample sizes in terms of population and medication, which improves its external validity. Moreover, the wide range of drugs studied may allow prioritization in future interventions.

The study presents some limitations that should be considered. First, noninitiation could have been underestimated because dispensed medication may not have been consumed and, conversely, it may have been overestimated if the patient had surplus medication from previous prescriptions or when prescriptions were pro re nata. Migrations could have caused an overestimation of the prevalence of 6-month noninitiation, although only 0.69% of children 0 to 14 years old emigrated in 2019.^46,47  Second, some variables that have been described as affecting adherence, such as nationality or medication beliefs,^35,48  were not available in the database. A sensitivity analysis extending the follow-up period to 6 months was conducted to quantify the possible impact of these variables. Variables related to the caregiver are likely to influence noninitiation in the pediatric population, although it should be pointed out that the children could not be linked to their caregivers in our database. Third, there could be some collinearity between copayment level and medication cost as well as between specialty of prescribing clinician and center where the prescription was issued. Sensitivity analyses were conducted to explore the impact of excluding one of these variables at a time (Supplemental Tables 10 and 11), and the interpretation of results was slightly affected. Fourth, this study was performed in a specific health care context, so the generalizability of the results to other countries should be done with caution because of possible differences within the health care system and the sociodemographic context. Fifth, the variable medication cost assumed by the patient showed different results on the basis of how it is modeled (Supplemental Tables 7 and 8); therefore, these results should be interpreted carefully. Finally, some data related to copayment and diagnoses were missing, and this can be understood as a consequence of the inherent limitations of working with real world data.

The prevalence of noninitiation of medical treatments in pediatrics varies according to patients’ age and medical groups and is noticeably high in Catalonia. Noninitiation rates of chronic medications diminished significantly from the 1- to 6-month follow-up. Factors related to the patient, such as age or copayment level, or to the health care system, such as type of prescriber or health care center, increase the risk of noninitiation. Results suggest that there are inequities in access to pharmacologic treatments in this population that must be taken into account by health care planners and providers. Physicians should be alert to noninitiation of chronic treatments, emphasizing the relevance of early adherence and resolving patients’ and parents’ doubts at the moment of prescription. Further research should be focused on understanding the root causes of noninitiation, including those related to the physician–patient relationship; and the clinical and economic impact of noninitiation in the pediatric population must be explored and axes of inequalities assessed.

We are grateful to Stephen Kelly for his contribution to English language editing.

ATC

Anatomical Therapeutic Chemical

CI

confidence interval

ICD-10

International Classification of Diseases, 10th Revision

OR

odds ratio

PC

primary care

SC

secondary care