Patient-tailored interventions to improve specialty medication adherence: results from a prospective randomized controlled trial

Abstract

Background:

Specialty medication nonadherence results in poor clinical outcomes and increased costs. This study evaluated the impact of patient-tailored interventions on specialty medication adherence.

Methods:

A pragmatic, randomized controlled trial was conducted at a single-center health-system specialty pharmacy from May 2019 to August 2021. Participants included recently nonadherent patients prescribed self-administered specialty medications from multiple specialty clinics. Eligible patients were stratified by historical clinic rates of nonadherence and randomized 1:1 to usual care or intervention arms. Intervention patients received patient-tailored interventions and 8 months of follow up. A Wilcoxon test was used to test the difference in 6, 8, and 12-month post-enrollment adherence, calculated using proportion of days covered (PDC), between the intervention and usual care arms.

Results:

There were 438 patients randomized. Baseline characteristics were similar between groups, mostly female (68%), white (82%), with a median age of 54 years (interquartile range [IQR] 40, 64). The most common reasons for nonadherence in the intervention arm were memory (37%) and unreachability (28%). There was a significant difference in median PDC between patients in the usual care and intervention arms at 8-months (0.88 vs. 0.94, p<0.001), 6-months (0.90 vs. 0.95, p=0.003) and 12-months post-enrollment (0.87 vs. 0.93, p<0.001).

Conclusions:

Patient-tailored interventions resulted in significant specialty medication adherence improvement compared to standard of care. Specialty pharmacies should consider targeting nonadherent patients for adherence interventions.

Introduction

Specialty medications are complex, high-cost therapies used to treat conditions such as inflammatory bowel disease, hormone deficiencies, and neurological disorders and have reduced morbidity and/or mortality in rheumatoid arthritis, cystic fibrosis , Hepatitis C infections and some cancers.^1–4 Specialty medication nonadherence can result in suboptimal therapeutic responses and significant specialty pharmacy financial losses from direct and indirect remuneration fees and value-based contract requirements.⁵ Identifying and addressing nonadherence could improve clinical outcomes and reduce overall healthcare costs.

Specialty medication adherence may be particularly challenging due to unique access, distribution, monitoring requirements, side effects and complicated administration techniques.^6–8 Specialty medications require complex third-party approvals and specialty tier copays that are often high, necessitating additional financial assistance. The clinical and administrative burden of specialty medications likely contribute to nonadherence across the spectrum of specialty disease areas.

Many health systems have developed integrated specialty pharmacies wherein pharmacists and technicians are embedded in clinics to manage specialty medications, including treatment selection, ensuring medication access and affordability, monitoring patient outcomes, and coordinating financial and clinical transitions of care throughout treatment.^{9, 10} Though health system specialty pharmacies (HSSP) have demonstrated high adherence rates in many disease states, more research is needed to identify, understand, and address patients who experience suboptimal adherence.^{9, 11–14} This study evaluated the impact of patient-tailored interventions on adherence in patients with recent nonadherence to specialty medications.

Materials and Methods

Study Design

A pragmatic randomized controlled trial was conducted to compare adherence between patients randomized to usual care and those receiving patient-tailored interventions and 8 months of follow up. The study was performed at an integrated HSSP in the southeast United States between May 2019 and August 2021. A detailed description of study design was previously published.¹⁵

Patients filling a self-administered specialty medication at the HSSP from an adult rheumatology, multiple sclerosis, lipid, adult miscellaneous, pulmonary or pediatric clinic having a proportion of days covered (PDC) <0.9 in the previous 4 months and 12 months for a single specialty medication at the 12-digit generic product identifier (GPI) level (a numerical system for identifying drugs with increasing specificity from 2 to 14 digits) were identified by an automated structured, query language-based report using specialty pharmacy claims data and imported into Research Electronic Data Capture (REDCap), a HIPAA-compliant, web-based data warehouse.^16,17 Patients were excluded if upon electronic health record (EHR) review they were misidentified as nonadherent (defined below), had a prescription from an outside provider, were deceased, planned to discontinue medication within the next 8 months or became incarcerated. Eligible patients were stratified by historical rates of nonadherence for the clinic from which the specialty medication was prescribed. (Appendix Table 1) and randomly assigned to either the usual care or intervention arm with 1:1 allocation using a randomization sequence generated by a statistician and deployed through REDCap’s randomization module. If a patient was randomized to the intervention arm but later identified to meet exclusion criteria, no interventions were provided, but the patient was included in the final analysis.

A patient focus group was conducted with six adult patients to identify reasons for nonadherence and recommended interventions for improving adherence. A pharmacist focus group was also conducted with six specialty pharmacists to understand pharmacists’ perception of reasons for patient nonadherence and effective interventions. Nonadherence reasons were categorized and literature on interventions for addressing each category was reviewed to create a list of potential interventions.¹⁵

The pragmatic study design was supported by the Vanderbilt Institute for Clinical and Translational Research Learning Healthcare System Platform which conducts research studies using a unique model that leverages pragmatic, randomized, controlled clinical trials embedded within usual care.¹⁸ The Vanderbilt University Institutional Review Board approved this clinical trial, and it was registered in ClinicalTrials.gov (NCT03709277) on October 17, 2018.

Setting (Usual Care)

In usual care, HSSP pharmacists and technicians worked in specialty clinics to manage specialty medications.⁹ Pharmacist assessments occurred quarterly to annually, depending on clinic protocol. During monthly refill calls, technicians asked patients for the number of and reasons for missed doses. Missed doses triggered pharmacist assessments that were addressed based on the pharmacists’ clinical judgement.

Interventions

Patients randomized to the intervention arm were contacted by a pharmacist and asked a series of semi-scripted questions (Appendix A) to obtain reasons for nonadherence Patient-reported reasons were combined with reasons identified through EHR review. Patients could report more than one reason for nonadherence. Interventions were provided on the screening call, or as soon as possible thereafter by either of two designated specialty pharmacists. After the initial intervention, the pharmacists followed up as needed (ranging from once to daily) until 8 months post-enrollment. If additional reasons for nonadherence were identified, another intervention was performed. The specialty pharmacist proactively reviewed medication orders to address any potential fulfillment barriers (e.g., refills available, copay assistance and prior authorization active) and scheduled refills. Table 1 lists common interventions for nonadherence reasons.

Table 1.

Intervention definitions and examples

Nonadherence Reason	Definition	Pharmacist intervention examples
Memory	Patient-reported challenges remembering to administer therapy as prescribed	• Educate on smartphone reminder set-up • Provide a daily pill box (oral medications) • Recommend adjustments in daily routine
Unreachable	Previous failure of at least 2 contact attempts to schedule a refill	• Call or message through a patient portal • Attempt to contact authorized alternative contact numbers • Mail letter sent to home address • When contact was successful, set up refills, update contact information and create a subsequent contact plan
Unresponsive	Lack of response to request to provide necessary labs, attend appointments, or obtain paperwork required to allow for safe or affordable continuation of treatment	• Contact patients and coordinate refill requirement completion (e.g., scheduling an appointment or obtaining labs)
Clinical	Medication administration or clinical response challenges	• Counsel on medication side effects • Provide educational materials
Social	Life circumstances that interfered with a patient’s ability to obtain or administer their medication	• Provide encouragement • Suggest social resources where available/known
Health literacy	Patient-reported lack of understanding the role of their specialty medication in their condition	• Counsel on medication purpose • Provide a one-page, plain language summary addressing frequently asked questions about commonly prescribed medications
Health-system Determinants	Prescriptions routed to a non-preferred pharmacy or a typo in the entry of prescription data resulting in delayed or missing refill tasks	• Errors corrected • Notify appropriate personnel to prevent the same error from reoccurring
Financial	Inability to afford medication usually due to a gap in financial assistance	• Assist with financial assistance renewals • Identify and discuss patient assistance options
No known reason	Inability to identify a reason for nonadherence or patient denial of any missed doses	• Counsel on the importance of adherence • Provide positive reinforcement

Study Endpoints

The primary outcome was PDC calculated at 8-months post-enrollment. Secondary outcomes included PDC at 6 and 12-months post-enrollment, reasons for nonadherence, and number of interventions performed.

Nonadherence Definitions and Calculations

Preliminary analysis revealed few patients serviced by the HSSP had a PDC <0.8; therefore, nonadherence was defined as PDC <0.9 for inclusion criteria to increase sample size. Misidentified nonadherence is used here for instances when a patient appeared to be nonadherent based on PDC, but EHR review revealed that treatment gaps were either clinically appropriate or expected due to external fills, and therefore not an opportunity to improve adherence.

PDC was calculated by generating a supply diary for the window of interest using raw refill data at the 8-digit GPI level. The supply diary calculated the day-to-day tally of medication available for each day in the window, where excess supply due to overlapping refills was shifted forward (but never backward) so only a one-day supply was available each day. Definitions and rationales for chosen PDC parameters for this study were published.¹⁵ Due to the lower specificity of the 8-digit GPI level used in analysis versus the 12-digit GPI level used in the database query for enrollment, few included patients had a retrospective 12-month PDC ≥0.9.

Statistical Analysis

A sample size of 438 patients was calculated based on a power of at least 90% at a type I error rate of 0.05, and the determination that an increase in PDC ≥ 5% would represent meaningful improvement in a population with high baseline adherence.¹⁵ Patient characteristics and pharmacist interventions were summarized using quartiles or mean and standard deviation (SD) for continuous variables and counts and proportions for categorical variables. Each patient’s PDC was calculated using raw refill data at the 8-digit GPI level. PDC values were ranked and treated as an ordinal outcome due to PDC having a skewed distribution and a range restricted between 0 and 1, violating the assumptions of linear regression models.¹⁹ The primary outcome was analyzed using intention-to-treat and included all randomized patients with at least two fills during the follow-up period. A Wilcoxon test was used to test the difference in 6, 8, and 12-month post-enrollment PDC between the intervention and usual care arms. An ordinal regression analysis was used to test the difference in 8-month PDC between patients in the intervention arm versus usual care arm, while controlling for the following covariates: retrospective 12-month PDC calculated prior to randomization, age, gender, race, insurance type, length of time on therapy (time between start date and date of enrollment), route of administration, clinic type, and patient portal status (having an active login allowing communication through the health system’s EHR). Analyses were conducted with the R programming language, version 4.0.2.²⁰

Results

Patient Population

Of the 1300 patients reviewed, 862 were excluded due to meeting at least one of the exclusion criteria, leaving 438 patients to be randomized. (Figure 1). Baseline characteristics were similar between groups (Table 2), mostly female (68%), white (82%), with a median age of 54 years (interquartile range [IQR] 40, 64). Patients were commonly from the adult rheumatology (35%) and multiple sclerosis (20%) clinics. Median retrospective 12-month PDC was 0.88 (IQR 0.78, 0.90) and 0.86 (IQR 0.78, 0.89) in the intervention and usual care arms, respectively.

Figure 1.

CONSORT diagram

Table 2.

Characteristics of randomized patients at baseline^a

Characteristic	Usual Care (N=220) % (n)	Intervention (N=218) % (n)	Combined (N=438) % (n)
Age, years- median (IQR)	54 [41, 65]	54 [40, 63]	54 [40, 64]
PDC in previous 12 months- median (IQR)	0.86 [0.78, 0.89]	0.88 [0.78, 0.90]	0.87 [0.78, 0.90]
Sex
Female	65.0 (143)	71 (155)	68 (298)
Male	35.0 (77)	29 (63)	32 (140)
Race
White	79.1 (174)	85.3 (186)	82.0 (360)
Black or African American	15.5 (34)	11.5 (25)	13.5 (59)
Other	1.8 (4)	1.8 (4)	1.8 (8)
Unknown	3.6 (8)	1.4 (3)	2.5 (11)
Insurance Type
Commercial	58.6 (129)	57.8 (126)	58.2 (255)
Medicare	35.5 (78)	34.4 (75)	34.9 (153)
Medicaid	5.9 (13)	7.8 (17)	6.8 (30)
Clinic Type
Adult Rheumatology	35.5 (78)	33.9 (74)	34.7 (152)
Multiple Sclerosis	17.7 (39)	21.6 (47)	19.6 (86)
Lipid	17.7 (39)	16.5 (36)	17.1 (75)
Adult Miscellaneousb	14.5 (32)	11.5 (25)	13 (57)
Pulmonaryc	8.2 (18)	8.7 (19)	8.4 (37)
Pediatricsd	6.4 (14)	7.8 (17)	7.1 (31)
Medication Route
Injectable	73.6 (162)	66 (145)	70 (307)
Oral	26.4 (58)	34 (73)	30 (131)
Active Patient Portal
Yes	66.4 (146)	69 (150)	68 (296)
No	33.6 (74)	31 (68)	32 (142)
Length of Therapy at Enrollment
< 1 year	35.5 (78)	30.7 (67)	33.1 (145)
≥1 year	63.6 (140)	69.3 (151)	66.4 (291)
Unknown	0.9 (2)	0 (0)	0.5 (2)

PDC = Proportion of Days Covered

^aContinuous variables are presented as medians with the interquartile range in brackets. Categorical variables are presented as percentages with sample size in parentheses. Percentages may not total to 100 due to rounding.

^bIncludes Adult Endocrinology, Dermatology, Hematology, Neurology and unassigned clinics

^cIncludes Pediatric cystic fibrosis , Adult cystic fibrosis , Asthma, IPF and PAH

^dIncludes Pediatric Irritable Bowel Disease and Pediatric Rheumatology

Reasons for Nonadherence

Of the 218 patients in the intervention arm, 297 reasons for nonadherence were identified. The most common reasons were memory (37%) and unreachability (28%). Many patients had no known reason (16%) or were unresponsive (14%). Few cited clinical or social issues (11%), health literacy (9%), health-system determinants (7%), or financial challenges (4%).

Intervention

A total of 352 interventions were provided with an average of 1.6 interventions per patient (SD = 1.4). Pharmacists made 1,469 attempts to contact patients with an average of 6.7 (SD = 5.6) attempts per patient including 503 successful phone calls, 751 unsuccessful phone calls, 96 written patient portal messages, 70 emails, and 49 letters mailed.

Adherence

After excluding patients with the less than 2 fills required to calculate PDC, 201 and 199 patients were included in the 8-month PDC analysis for the intervention and usual care cohorts, respectively. There was a significant difference in median PDC between patients in the intervention and usual care arms at 8-months (0.94 [IQR 0.84, 1.0] vs. 0.88 [IQR 0.75, 0.97], p<0.001). A significant difference also existed at 6-and 12-months post-enrollment (0.95 [IQR 0.84, 1.0] vs. 0.90 [0.76, 0.98], p=0.003; 0.93 [IQR 0.82, 0.98] vs. 0.87 [IQR 0.72, 0.95] p<0.001, respectively). Figure 2 shows the trend in PDC over time.

Figure 2.

Proportion of days covered (PDC) over time

The violin plot visualizes the distribution of PDC in the usual care and intervention arms at several time points across the study. A trend line connect the median values for each treatment group at each PDC timeframe. m indicates months

Patients in the intervention arm were 1.8 times more likely to have a higher PDC than patients in the usual care arm at 8 months after adjusting for patient factors that might affect PDC (odds ratio [OR] 1.8; 95% confidence interval [CI] 1.3 to 2.6; p=0.001) (Figure 3).Clinic type (p=0.025) and higher retrospective 12-month PDC (OR 1.4, CI 1.2 to 1.7; p<0.001) were also significantly associated with 8-month post-enrollment PDC Intervention patients from the pediatric and adult rheumatology clinics had the highest observed difference in median 8-month post-enrollment PDC over their usual care arm counterparts; whereas, Lipid and Multiple sclerosis clinic patients in the control arm had higher mean PDC values than the intervention patients at 8-months post-enrollment (Appendix Figure 1).

Figure 3.

Ordinal logistic regression of 8-month post-enrollment PDC

The forest plot shows the results of the ordinal logistic regression analysis for PDC at 8 months after enrollment. The horizontal lines extending from each point indicate the 95% confidence interval of each odds ratio.

Discussion

This pragmatic randomized controlled study demonstrated that patient-tailored adherence interventions targeting patients with recent nonadherence can significantly improve specialty medication adherence. The odds of having higher PDC at 8 months was 1.8 times higher in patients who received interventions than those receiving usual care. Despite a high median retrospective PDC of 0.88 in the intervention arm, many patients benefited from patient-tailored interventions. Specialty medication nonadherence can lead to disease progression, reduced productivity and quality of life, and increased healthcare utilization, making these results significant for patients, providers, specialty pharmacies, and payors.^21–25 Patient-tailored interventions had positive early and long-term effects, suggesting these interventions can help patients become more adherent quickly, and sustain better adherence behaviors up to 4 months post-intervention. Previous studies have demonstrated an improvement in adherence following HSSP integration, or compared to non-HSSPs, in oncology and Human Immunodeficiency Virus treatment.^{12, 14, 26}

Previous studies evaluating nonadherence to specialty medications have focused on specific specialty disease states and relied on retrospective review to identify nonadherence reasons. The current study prospectively evaluated nonadherence using both chart review and patient-reported adherence assessments. The most common reason for nonadherence was memory, followed by being unreachable for a refill. Memory aids such as text messages, alarms, beeper pill caps, and telephone calls have demonstrated varying results on improving adherence rates.^27–29 For patients with memory challenges, specialty pharmacists in the current study provided patients with memory aids and educated patients on smartphone reminders. Because the integrated HSSP schedules medication refills only after communicating with the patient (via phone or EHR portal), being able to reach the patient is essential. Pharmacists contacted patients via multiple modalities, including mailers, which have increased adherence in non-specialty disease states, then created comprehensive communication plans including preferred method and alternate contacts.²⁷ HSSPs should consider having memory aid discussions and establishing comprehensive communication plans into their initial medication counseling to prevent nonadherence.

Retrospective 12-month PDCs and clinic type were associated with prospective PDC rates. It is not surprising that patients with a higher baseline PDC would also have a higher 8-month post-enrollment PDC. Adherence in the lipid and multiple sclerosis clinics may have been impacted by environmental factors. Both specialty medications in the Lipid clinic had significant price reductions. The multiple sclerosis clinic had an additional pharmacist focused on assisting with care coordination during the study period. High 8-month post-enrollment PDC was seen in patients from the pediatric and adult rheumatology clinics in the intervention compared to usual care. Previous studies found that medication-related knowledge and medication adherence are low in adolescent patients with inflammatory bowel disease .³⁰ Studies of children with cancer entering adolescence have shown that increase in medication administration variability, such as non-daily dosing regimens which are common for specialty medications, can lead to more nonadherence.^{31, 32} The pharmacists in this study provided targeted interventions to assist with these potential adherence challenges such as health literacy and memory that likely contributed to the increased PDC in the pediatric population. Adult rheumatology patients have been shown to be nonadherent due to adverse effects, perceived treatment inefficacy, out-of-pocket costs, and medication fulfillment logistics.^{13, 33, 34} Targeted interventions addressing health literacy and financial and medication fulfillment challenges likely contributed to the large difference in 8-month prospective median PDC between intervention and usual care arms in the adult rheumatology clinic. Specialty pharmacies could target patients from specific clinics for nonadherence to maximize the impact of added adherence services.

Implications for Practice

Several implications for practice were noted in this study. PDC from claims alone resulted in a high rate of misidentified nonadherence (60%). Specialty pharmacies likely need additional clinical data to assess true adherence behavior, and these limitations should be considered when utilizing PDC as a metric for specialty pharmacy practice. HSSPs have access to the EHR and are uniquely positioned to evaluate whether gaps in therapy may be appropriate. Most interventions did not require the clinical or disease state expertise at the level of a pharmacist, suggesting other roles such as advanced pharmacy technicians, case managers/social workers, or nurses could be used to perform many interventions to address common adherence challenges.

Limitations

Though the prospective, randomized study design strengthens the findings, limitations exist. The intervention arm was not always blinded because clinic-based pharmacists and providers were occasionally informed of interventions and had visibility into EHR documentation. A priori, all pharmacists and technicians were informed about the study and asked not to alter care based on knowledge of enrollment. No indication of assignment was provided to patients or documented in the patient record. Reasons for nonadherence could not be confirmed by patients in the usual care arm, which may have resulted in patients being enrolled in the usual care arm that should have been excluded. Factors that affect adherence are abundant (e.g., polypharmacy, comorbidities, social determinants of health); due to sample size, we limited the number of covariates included in the model to those determined to have the largest potential impact on adherence. As with all studies using pharmacy claims to calculate adherence, PDC does not evaluate medication administration, rather medication refill behavior. Generalizability to non-health system specialty pharmacies may be limited.

Conclusions

Study findings suggest that targeting nonadherent patients, identifying their reason(s) for nonadherence, and providing patient-tailored interventions are associated with increased adherence to specialty medications. Specialty pharmacies should consider designing adherence intervention services to target nonadherent patients. More studies are needed to determine the clinical and financial impact of specialty medication adherence interventions.