1 Study Details

1.1 Study title

Pharmacist Interventions to Improve Specialty Medication Adherence: A Randomized Controlled Trial

1.2 Background

The effectiveness of specialty medications in complicated clinical conditions depends on adherence to therapy. Patient-tailored adherence interventions may be able to improve adherence in non-adherent patients.

1.3 Objectives

Primary Objective

  • Assess effectiveness of patient-tailored intervention on improving 8-month PDC

Secondary Objectives

  • Describe the rates and reasons for non-adherence

1.4 Methods

  • Single center, pragmatic, randomized controlled trial of usual care compared with usual care plus patient-tailored adherence interventions.

  • Patients were enrolled May 10, 2019 through August 10, 2021

  • Patients randomized to the intervention arm undergo a baseline assessment to clarify reasons for non-adherence and subsequently receive patient-tailored interventions based on their specific reasons

  • 438 patients enrolled and trial will be ongoing from 10 May 2019 to the first half of 2020 with 8 months of follow-up for each patient.

  • Primary outcome: PDC calculated at 8 months post-enrollment

1.4.1 Exclusion Criteria

  • PDC > 90% in previous 4 months or previous 12 months.

The PDC for the exclusion criteria was calculated in order to enrich the sample for non-adherent patients. The baseline PDC secondary outcome was calculated independently and uses the same algorithm as used to calculated the prospective PDC outcome. Due to the slightly different methods, some patient’s baseline PDCs are greater than 90%.

1.5 Outcomes

  • 6, 8, and 12 month post-enrollment follow-up PDC

  • 12 month pre-enrollment baseline PDC

  • Baseline Assessment Results

1.6 PDC Calculations

  • Medication adherence is estimated by calculating the proportion of days covered (PDC) from participants’ fill data from the electronic healthcare record.

    • Unlike other adherence estimates such as medication possession ratio, PDC accounts for oversupply and cannot be greater than 100%

    • A PDC of 1 indicates perfect adherence while a PDC of 0 indicates no use of medication.

    • \(PDC=\frac{\text{Sum of days covered in timeframe}}{\text{Number of days in time frame}}\)

      • Days covered within a time frame is a term used to indicate availability of a prescribed daily dose on a given day.

      • The availability of medication is tracked using a moving array of each day’s supply that accounts for overfills by moving extra medication within a fill period to the start of the first exhaustion date of the given medication

      • Oversupply is truncated at the end of the time frame and each day in the denominator is limited to have a maximum of 1 day’s supply.

  • For this study, the time frame is defined as the time period from the start of the first fill to the first day of the last fill.

  • Three different time-frames were specified for PDC calculations in this analysis

    • 12 month retrospective baseline PDC: The time frame starts at the first medication fill within 1 year before trial enrollment and ends at the start of the last fill prior to trial enrollment

    • 6, 8, and 12 months prospective PDC: The time frame starts at the first medication fill after trial enrollment and ends at the start of the last fill before 6, 8, or 12 months after trial enrollment.

2 Descriptive Statistics

2.1 Demographics and Characteristics

Demographics and characteristics
N
N=439
Age 439 40 53 64  (51 ±18)
Sex 439
    Male 32% (140)
    Female 68% (299)
Race 439
    White 82.0% (360)
    Black or African American 13.4% ( 59)
    Other 2.1% ( 9)
    NA 2.5% ( 11)
Insurance 439
    Commercial 58.1% (255)
    Medicare 34.9% (153)
    Medicaid 7.1% ( 31)
    None 0.0% ( 0)
    Other 0.0% ( 0)
Medication Route 439
    Injectable 70% (307)
    Oral 30% (132)
Duration of Therapy 439
    < 1 year 33.0% (145)
    >= 1 year 66.5% (292)
    NA 0.5% ( 2)
Clinic 439
    Adult Misc 13.0% ( 57)
    Adult Rheumatology 34.6% (152)
    Lipids 17.1% ( 75)
    MS 19.6% ( 86)
    Pediatric 7.1% ( 31)
    Pulmonary 8.7% ( 38)
Clinic Level 439
    High 25.3% (111)
    Medium 33.3% (146)
    Low 41.2% (181)
    NA 0.2% ( 1)
Patient portal use 439
    No 33% (143)
    Yes 67% (296)
Baseline PDC 12m 439 0.78 0.87 0.90  (0.81 ±0.12)
a b c represent the lower quartile a, the median b, and the upper quartile c for continuous variables. x ± s represents X ± 1 SD.   N is the number of non-missing values. Numbers after proportions are frequencies.

2.2 Social Determinants of Health

  • SDOH data is estimated using geocoding and American Census Survey data
Social determinants of health
N
N=439
Distance to Care 437 22 51 112  ( 85 ±123)
Percent Poverty 437 6.9 12.0 19.6  (13.8 ± 9.2)
Percent No Health Insurance 437 5.6 9.1 13.4  (10.0 ± 5.6)
Median Income 437 4.2 5.3 7.6  (6.2 ±3.0)
Percent Highschool Education 437 82.6 88.7 94.5  (88.2 ± 7.4)
Percent Assisted Income 437 5.1 11.0 16.7  (12.4 ± 9.2)
Deprivation Index 437 0.24 0.32 0.38  (0.31 ±0.10)
Percent Vacant 437 5.0 8.4 13.4  (10.2 ± 8.4)
a b c represent the lower quartile a, the median b, and the upper quartile c for continuous variables. x ± s represents X ± 1 SD.   N is the number of non-missing values.

3 Reasons for Non-adherence

The reasons for the non-adherence for the patients that participated in the randomized trial and the reasons for the misidentified non-adherent patients (none of which are in the RCT data) are displayed below.

3.1 Non-adherent Patients

3.1.1 Total Occurrences - Non adherent

The table gives the total number of occurrences for each non-adherence reason, but does not account for patients having intersections of multiple reasons.

True non-adherent: Total count and percent of selected reasons
Count (n= 299)
Financial 8 (2.7%)
Clincial 25 (8.4%)
Health Literacy or comprehension 19 (6.4%)
Memory 82 (27.4%)
Unreachable 60 (20.1%)
Unresponsive 32 (10.7%)
Pharmacy/Clinic Error 15 (5%)
Social Issues 23 (7.7%)
No known reasons 35 (11.7%)

3.1.2 Upset plot - Non adherent

The upset plots are an alternative to a venn diagram and can be helpful when a set size is large. This upset plot gives the total number of occurrences for each non-adherence reason in the bottom left corner of the figure and is labeled “set size”. The “intersection size” graph on the top of the figure indicates the total number of occurrences for every intersection between the different reasons. The intersection is indicated by the dots at the bottom portion of the figure. The figure is not exhaustive and excludes low count intersections.

3.2 Misidentified Patients

775 patients were flagged as non-adherent when there was a reason for their non-adherence - these are referred to as “misidentified patients”. Two of misidentified patients were randomized into trial, but they were randomized before the cutoff date and were excluded. The following tables and charts summarize the count and patterns of the misidentified non-adherence reasons.

3.2.1 Total Occurences - Misidentified

The table gives the total number of occurrences for each non-adherence reason, but does not account for patients having intersections of multiple reasons.

Misidentified non-adherent: Total count and percent of selected reasons
Count n=(775)
Clinical 250 (32.3%)
Non-VUMC fill 136 (17.5%)
Medication discontinuation 167 (21.5%)
Drug dosage form incongruency 29 (3.7%)
Temporary insurance laspse 54 (7%)
Switched to alternate GPI of same med 87 (11.2%)
No longer seeing VUMC 16 (2.1%)
Held for pandemic 9 (1.2%)
Other 45 (5.8%)
Delay in starting therapy 1 (0.1%)
Out of country 0 (0%)

3.2.2 Upset plot - Misidentified

The upset plots are an alternative to a venn diagram and can be helpful when a set size is large. This upset plot gives the total number of occurrences for each non-adherence reason in the bottom left corner of the figure and is labeled “set size”. The “intersection size” graph on the top of the figure indicates the total number of occurrences for every intersection between the different reasons. The intersection is indicated by the dots at the bottom portion of the figure. The figure is not exhaustive and excludes low count interestions.

3.3 Reasons for non-adherence by clinic

3.3.1 Barplot

3.3.2 Heatmap

4 Randomized Pharmacist Intervention

4.1 Descriptive Statistics

4.1.1 Interventions

A total of 355 interventions were performed on 190 patients (29 out of the 219 patients in the intervention group did not have interventions). The distribution of the number of interventions per patient in the intervention group is displayed below.

Interventions per patient
N
N=219
Intervention Count 219 1.0 1.0 2.0  (1.6 ±1.4)
a b c represent the lower quartile a, the median b, and the upper quartile c for continuous variables. x ± s represents X ± 1 SD.   N is the number of non-missing values.

4.1.2 Contact attempts

Contacts per patient in treatment group
N
N=219
Contacts per patient 219 2.0 5.0 10.0  ( 6.7 ± 5.6)
a b c represent the lower quartile a, the median b, and the upper quartile c for continuous variables. x ± s represents X ± 1 SD.   N is the number of non-missing values.

Total count of each contact type
N
N=1475
Contact type 1475
    Mailed letter 3.3% ( 49)
    Phone call- Left voicemail 34.0% (502)
    Phone call- No voicemail or contact 17.1% (252)
    Phone call- Spoke to patient 34.2% (504)
    Sent email 4.8% ( 71)
    Sent MHAV message 6.6% ( 97)
N is the number of non-missing values. Numbers after proportions are frequencies.

4.2 Outcomes

4.2.1 Post-enrollment PDC

4.2.1.1 Table

Prospective PDC Measurements
N
 Usual Care
N=220		 Intervention
N=219		 P-value
Baseline PDC 12m 439 0.78 0.86 0.89  (0.81 ±0.12) 0.78 0.87 0.90  (0.82 ±0.12) 0.21
Prospective 6mo PDC 397 0.76 0.90 0.98  (0.84 ±0.17) 0.84 0.95 1.00  (0.90 ±0.13) 0.003
Prospective 8mo PDC 401 0.75 0.88 0.97  (0.83 ±0.17) 0.84 0.94 0.99  (0.89 ±0.13) <0.001
Prospective 12mo PDC 402 0.72 0.87 0.95  (0.82 ±0.17) 0.82 0.93 0.98  (0.88 ±0.13) <0.001
a b c represent the lower quartile a, the median b, and the upper quartile c for continuous variables. x ± s represents X ± 1 SD.   N is the number of non-missing values.
Test used: Wilcoxon test .

4.2.1.2 Density plots

5 8 Month PDC Outcome

5.1 Univariate

Prospective 8 Month PDC
N
 Usual Care
N=220		 Intervention
N=219		 P-value
Prospective 8mo PDC 401 0.75 0.88 0.97  (0.83 ±0.17) 0.84 0.94 0.99  (0.89 ±0.13) <0.001
a b c represent the lower quartile a, the median b, and the upper quartile c for continuous variables. x ± s represents X ± 1 SD.   N is the number of non-missing values.
Test used: Wilcoxon test .

5.2 Ordinal logistic regression

5.2.1 Visualization

5.2.2 Table

95% Confidence interval
Variable Odds Ratio Lower Upper p-value
Treatment group (Usual care reference)
Intervention 1.82 1.28 2.59 <0.001
12 Month Baseline PDC
Per 10 unit increase 1.43 1.23 1.65 <0.001
Age
Per 10 year increase 1.13 0.97 1.31 0.111
Administration method (Injectable reference)
Oral 1.25 0.78 2.01 0.361
Sex (Male reference)
Female 1.07 0.71 1.60 0.746
Race (White reference)
Black or African American 1.50 0.88 2.56 0.134
Other 2.92 0.82 10.36 0.098
Duration of treatment (Less than one year reference)
Greater than or equal to 1 yearr 1.24 0.83 1.85 0.302
Insurance (Commercial reference)
Medicare 0.97 0.62 1.52 0.901
Medicaid 1.22 0.50 2.97 0.661
Clinic (Rheumatology reference)
Adult Miscellaneous 1.61 0.91 2.87 0.104
Lipids 2.11 1.21 3.67 0.008
Multiple Sclerosis 2.15 1.24 3.70 0.006
Pediatric 2.09 0.75 5.84 0.158
Pulmonary 1.35 0.65 2.82 0.420
Online portal use (Inactive reference)
Active 1.13 0.76 1.67 0.545

5.3 8 month PDC by treatment group and clinic

5.3.1 Median difference in PDC pre and post-intervention by clinic

Median difference in Post/Pre PDC by treatment group and clinic
Clinic Usual Care Intervention
Adult Rheumatology 0.02 (n=78) 0.07 (n=74)
Adult Misc 0.03 (n=32) 0.09 (n=25)
MS 0.09 (n=39) 0.1 (n=47)
Lipids 0.08 (n=39) 0.09 (n=36)
Pulmonary 0.05 (n=18) 0.09 (n=20)
Pediatric 0.04 (n=14) 0.07 (n=17)
All 0.06 0.09

5.4 PDC by timeframe and clinic

5.4.1 Usual Care

5.4.2 Intervention

5.4.3 Overall

5.5 PDC by reason for non-adherence in Intervention group