Impact of pharmacist-physician collaborative care on hemoglobin A1c and blood pressure quality measure achievement in primary care

Abstract

BACKGROUND:

Multidisciplinary primary care models incorporating pharmacists have emerged to improve glycemic control in patients with uncontrolled type 2 diabetes mellitus (T2DM). Healthcare Effectiveness Data and Information Set (HEDIS) measures establish quality benchmarks for comprehensive diabetes care and guide reimbursement. Large-scale research on the effect of pharmacist interventions to improve these quality measures in primary care remains limited.

OBJECTIVE:

To evaluate the effectiveness of a pharmacist-physician collaborative care (PPCC) model on comprehensive diabetes care quality measure achievement compared with standard care (SC).

METHODS:

This retrospective cohort study included adults aged 18 to 75 years with uncontrolled T2DM receiving care in primary care clinics at a community-based health system in Virginia from July 1, 2018, to December 31, 2019. Patients were in one of 2 groups: (1) the intervention group (PPCC), where embedded pharmacists provided diabetes management under a collaborative practice agreement, and (2) the comparator group receiving SC in clinics without pharmacists. The SC group was created via 1:2 propensity score matching. Generalized linear mixed models assessed the association between group and quality measure achievement. Primary outcomes included glycated hemoglobin (hemoglobin A1c) (≤9%, ≤8%, ≤7%) and blood pressure control (<140/90 mm Hg), per the last recorded value in 2019.

RESULTS:

The sample (N = 1,293) had a mean age of 57 years, was 56% female, and 45% each White and Black. The PPCC group (n = 431) was more likely to achieve A1c control compared with the SC group (n = 862) (A1c <9%: odds ratio [OR] = 3.68, 95% CI = 2.31-5.84; A1c <8%: OR = 3.53, 95% CI = 2.12-5.89; A1c <7%: OR = 4.61, 95% CI = 2.48-8.56; all P < 0.01). Similarly, the PPCC group was more likely to achieve blood pressure control less than 140/90 mm Hg (OR = 1.49, 95% CI = 1.01-2.22; P = 0.04).

CONCLUSIONS:

Patients in the PPCC group were more likely to meet comprehensive diabetes care quality measures compared with SC. These results underscore the value of pharmacists in diabetes management in primary care and their contribution to value-based care.

Plain language summary

This study compared 2 ways of treating patients with hard-to-control type 2 diabetes and high blood sugar. Standard care (SC) from clinicians was compared with care in which pharmacists and clinicians collaborated to manage diabetes. Overall, 1,293 patients were included (July 1, 2018, to December 31, 2019). Results suggest that patients with both pharmacist and clinician care were 3 to 5 times more likely to achieve healthier blood sugar levels and had better blood pressure control than those receiving SC.

Implications for managed care pharmacy

Managed care organizations should consider integrating pharmacists into primary care diabetes management teams. Patients receiving pharmacist care were 3 to 5 times more likely to achieve hemoglobin A1c control than SC. This model significantly improves key quality measures that affect reimbursement, including glycated hemoglobin and blood pressure control. With pharmacists providing care under collaborative practice agreements, health systems could improve patient outcomes while meeting Healthcare Effectiveness Data and Information Set (HEDIS) benchmarks, potentially increasing value-based payments and reducing long-term complications from uncontrolled diabetes.

Type 2 diabetes mellitus (T2DM) affects approximately 38 million people in the United States, posing a significant public health challenge. ¹ T2DM is associated with an increased risk of acute and long-term complications, leading to substantial morbidity, mortality, and health care costs. ^{1 , 2} Despite widely available evidence-based clinical practice guidelines for diabetes care, many patients with T2DM fail to achieve recommended glycemic control and blood pressure (BP) targets. ^{3 – 5}

To evaluate and improve the quality of diabetes care, the National Committee for Quality Assurance established the Healthcare Effectiveness Data and Information Set (HEDIS) measures for comprehensive diabetes care. Adopted by more than 90% of US health plans covering 227 million people, ⁶ HEDIS measures include annual glycated hemoglobin (hemoglobin A1c) testing, A1c control, and BP control. ^{3 , 7} These quality measures are endorsed by the National Quality Forum ⁸ and used by the Centers for Medicare & Medicaid Services, ⁹ serving as standardized benchmarks for assessing the quality of diabetes care, identifying improvement opportunities, and monitoring quality initiatives. In the current value-based payment landscape, these measures are used to assess a health care organization’s overall effectiveness in delivering patient care.

In response to a projected shortage of primary care providers ¹⁰ and a demographic shift toward an older, more medically complex population, ^{11 , 12} many health care systems have adopted multidisciplinary care models that include pharmacists to help manage complex, medication-intensive conditions, such as hypertension and diabetes. These multidisciplinary models improve patient outcomes and increase the quality of care delivery in chronic disease management. ^{5 , 13 – 21} Whereas physicians typically spend less than a minute discussing new prescriptions during office visits, ²² pharmacists provide medication management and counseling through regular follow-up visits. ^{23 – 25} Studies examining pharmacists’ impact on quality measures have shown promise but often presented limitations such as imbalanced patient characteristics, single-arm designs without a comparator group, and small patient cohorts. ^{20 , 26 – 28} Additionally, these studies often used quality measures primarily as ad hoc indicators of glycemic and BP control rather than as intended in clinical practice for end-of-year patient outcome assessments. ^{26 , 29} Although previous research demonstrates the value of pharmacists in primary care, large-scale research evaluating their real-world impact on quality measure achievement remains limited.

The aim of this study was to evaluate the effectiveness of a pharmacist-physician collaborative care (PPCC) model on comprehensive diabetes care quality measure achievement for glycemic and BP control compared with standard care (SC) in a community-based health system.

Methods

STUDY DESIGN AND SETTING

This retrospective, observational cohort study used electronic health record (EHR) data from primary care clinics at Bon Secours Mercy Health (BSMH), in Virginia. The study compared outcomes among patients in the PPCC and SC groups. In the PPCC clinics, pharmacists provided diabetes management services under a collaborative practice agreement, which enabled them to initiate, modify, or discontinue diabetes medications, antihypertensives, and related therapies and order and interpret laboratory tests.

Eight primary care clinics with embedded pharmacists were matched to 5 comparator clinics without pharmacists based on clinic (i.e., panel size per full-time equivalent, proportion of uninsured and female patients) and socioeconomic characteristics of their geographic area (i.e., population density, median household income, race, median age, educational attainment, poverty level, and unemployment rate). This process was carried out by 3 of the authors, including the Director of Ambulatory Practice and Medication Management, who had deep knowledge about the network of clinics, their geographic location, and the population they served. At the patient level, propensity score matching was used to match patients on demographic and clinical characteristics, and diabetes-related factors known to influence diabetes outcomes, including age, sex, race, health insurance status, Charlson Comorbidity Index (CCI) score, number of chronic medications prescribed, and A1c baseline levels. ^{30 , 31} Each PPCC patient was matched to 2 SC patients based on the closest propensity scores without replacement (greedy nearest neighbor, with a caliper width of 0.25).

This study was deemed exempt by the Virginia Commonwealth University and BSMH Institutional Review Boards.

STUDY POPULATION

Eligible patients were aged 18 to 75 years with a diagnosis of T2DM and A1c greater than or equal to 7%. This age range was selected to match reporting criteria for HEDIS quality measures. ⁷ Patients were excluded if they did not have a diagnosis of diabetes or had a baseline A1c less than 7%. Data were collected from July 1, 2018, to December 31, 2019, covering the primary study period (2019) and the preceding 6 months. The inclusion of data from the 6 months prior to 2019 enabled the establishment of baseline laboratory values, comorbidities, and medication profiles for each patient. In the PPCC group, all new patients referred to the pharmacist for diabetes management in 2019 were included, with their first PPCC clinic visit serving as the index date. Patients were primarily referred to the pharmacist if they had uncontrolled diabetes (A1c ≥7%), with follow-up frequency individualized according to medication changes and safety monitoring needs. The typical follow-up happened every 1 to 3 months until glycemic control was achieved, after which visits became less frequent. In the SC group, new patients presenting to clinics without a pharmacist who had an A1c measurement greater than or equal to 7% during 2019 were included. The date of the first A1c measurement greater than or equal to 7% served as the patient’s index date.

MEASURES

Health care groups negotiate incentives based on quality measures differently. In this analysis, we focused on the HEDIS comprehensive diabetes care quality measures, applicable as of 2019, to increase the relevance of our findings across different health care settings. The primary outcomes selected for this analysis were the achievement of A1c control defined as less than 9.0%, less than 8.0%, and less than 7.0%; poor A1c control (>9.0%); and BP control (<140/90 mm Hg). ⁷ Achievement of these measures was assessed using the last recorded value during the 2019 calendar year.

The CCI score, a weighted index considering the number and severity of comorbid diseases, was calculated for each patient based on the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) diagnostic codes present at their index date. ³² Chronic medications were considered those taken daily for 30 or more days or used on an “as needed” basis for more than 6 months cumulatively within the past 12 months per EHR documentation. The total number of medications at the index date was used as a measure of medication burden. Total ambulatory care visits were counted as the number of outpatient encounters at any BSMH facility between the patient’s index date and December 31, 2019. The quarter of the year was categorized based on the patient’s index date (Q1: January to March, Q2: April to June, Q3: July to September, Q4: October to December) to assess a potential temporal relationship between the timing of patient visits and meeting quality measures. These covariates, along with demographic (i.e., age, sex, race, health insurance type, smoking status) and clinical characteristics (i.e., baseline systolic BP [SBP], baseline diastolic BP, hypertension diagnosis) were incorporated into the models to control for potential confounding and to identify sociodemographic disparities in quality measure achievement.

STATISTICAL ANALYSIS

Logistic regression models were used to generate propensity scores, which represent the probability of receiving the PPCC. Balance between groups was assessed using standardized mean differences, with values less than 0.1 considered indicative of good balance. All matched covariates were assessed using independent t-tests and chi-square tests to compare continuous and categorical variables, respectively.

To evaluate differences in quality measure achievement, the proportion of individuals achieving each measure in the PPCC and SC groups was compared using chi-square tests. Additionally, generalized linear mixed models with a logit link function were performed to account for the clustering of patients within clinics by including clinic-specific random effects. The generalized linear mixed models included study group, demographics (i.e., age, sex, race, health insurance type, smoking status), hypertension diagnosis, CCI score, number of chronic medications prescribed, number of ICD diagnosis codes, index date quarter, total ambulatory care visits, and baseline values of A1c, SBP, and diastolic BP as fixed effects. Adjusted odds ratios were calculated with 95% CIs. All analyses were performed using SAS version 9.4 (SAS Institute Inc.), with a 2-sided P value <0.05 considered statistically significant.

Results

At the beginning of the study, 2,664 patients met the inclusion criteria: 464 in the PPCC group and 2,200 in the SC group. Following propensity score matching (1:2), 33 patients from the PPCC group were excluded owing to lack of suitable matches in the SC group based on our predefined matching criteria, resulting in 431 patients in the PPCC group matched to 862 in the SC group. The sample (N = 1,293) had a mean age of 57.2 years; more than half were female (55.8%), 45.3% were White, and 45.3% were Black. Patients were primarily covered by commercial insurance (37.7%) or Medicare (36.3%) (Table 1). The groups were well balanced with regard to age, sex, race, insurance type, and number of chronic medications at baseline. The PPCC group presented with a significantly higher mean CCI score (1.5 vs 1.4; P = 0.03). No significant differences in A1c levels existed between groups at baseline (9.9% vs 9.7%; P = 0.08). Baseline characteristics of the sample before and after propensity score matching are presented in Supplementary Tables 1 and 2 ^{(113.3KB, pdf)} (available in online article).

TABLE 1.

Baseline Characteristics of Eligible Participants

Characteristic	Overall (N = 1,293)	PPCC group (n = 431)	SC group (n = 862)	P value
Demographics
Age, mean (SE), years	57.2 (0.38)	57.5 (0.65)	57.0 (0.47)	0.62
Sex, n (%)
Female	721 (55.8)	237 (55.0)	484 (56.1)	0.69
Male	572 (44.2)	194 (45.0)	378 (43.9)
Race, n (%)				0.77
White	586 (45.3)	195 (45.2)	391 (45.4)
Black	586 (45.3)	199 (46.2)	387 (44.9)
Other	121 (9.4)	37 (8.6)	84 (9.7)
Smoking status, n (%)				0.02 a
Current	167 (12.9)	55 (12.8)	112 (13.0)
Former	369 (28.5)	155 (36.0)	214 (24.8)
Never	641 (49.6)	214 (49.7)	427 (49.5)
Missing	116 (9.0)	7 (1.6)	109 (12.7)
Insurance
Medicare, n (%)	469 (36.3)	159 (36.9)	310 (36.0)	0.74
Commercial, n (%)	487 (37.7)	160 (37.1)	327 (37.9)	0.78
Other, n (%)	337 (26.0)	112 (26.0)	225 (26.1)	0.98
Number of ambulatory visits, mean (SE)	3.1 (0.07)	2.3 (0.11)	3.5 (0.08)	<0.01 a
Time of index visit				<0.01 a
Q1 (January to March), n (%)	578 (44.7)	150 (34.8)	428 (49.7)
Q2 (April to June), n (%)	335 (25.9)	122 (28.3)	213 (24.7)
Q3 (July to September), n (%)	206 (15.9)	81 (18.8)	125 (14.5)
Q4 (October to December), n (%)	174 (13.5)	78 (18.1)	96 (11.1)
Clinical
Charlson Comorbidity Index score, mean (SE)	1.4 (0.02)	1.5 (0.03)	1.4 (0.03)	0.03 a
Number of ICD-10 diagnoses, mean (SE)	7.2 (0.16)	3.1 (0.15)	9.3 (0.19)	<0.01 a
Number of medications, mean (SE)	9.8 (0.15)	10.0 (0.24)	9.6 (0.19)	0.22
Hypertension, n (%)	709 (54.8)	99 (23.0)	610 (70.8)	<0.01 a
Diabetes, n (%)	431 (100.0)	431 (100.0)	862 (100.0)	—
Baseline values
A1c, mean (SE)	9.8 (0.06)	9.9 (0.09)	9.7 (0.08)	0.08
SBP, mean (SE)	132.9 (0.52)	130.9 (0.79)	133.9 (0.67)	<0.01 a
DBP, mean (SE)	79.3 (0.31)	78.0 (0.49)	79.9 (0.38)	<0.01 a

^a P value is statistically significant.

A1c = glycated hemoglobin; DBP = diastolic blood pressure; ICD-10 = International Classification of Diseases, Tenth Revision; PPCC = pharmacist-physician collaborative care; SBP = systolic blood pressure; SC = standard care; SE = standard error.

In the bivariate analysis, the proportion of patients achieving A1c less than 9% and less than 7% was significantly higher in the PPCC group compared with the SC group, but no significant difference was observed for A1c less than 8%. The proportion of patients with uncontrolled diabetes (A1c >9%) was significantly lower in the PPCC compared with the SC group, and a significantly higher proportion of patients in the PPCC group achieved the target BP value compared with SC (Table 2).

TABLE 2.

Comparison of Diabetes Quality Measure Achievement (Bivariate Analysis)

Quality measure	PPCC group (n = 431)	SC group (n = 862)	P value
A1c control <9.0%, n (%)	276 (64.0)	491 (57.0)	0.01 a
A1c control <8.0%, n (%)	196 (45.5)	347 (40.3)	0.07
A1c control <7.0%, n (%)	82 (19.0)	95 (11.0)	<0.01 a
A1c poor control >9.0%, n (%) b	145 (33.6)	357 (41.4)	<0.01 a
BP control <140/90 mm Hg, n (%)	322 (74.7)	583 (67.6)	<0.01 a

^a P value is statistically significant.

^b Inverse relationship—lower frequency is better.

A1c = glycated hemoglobin; BP = blood pressure; PPCC = pharmacist-physician collaborative care; SC = standard care.

In the generalized linear mixed models, patients in the PPCC group were 3.7, 3.5, and 4.6 times more likely to achieve A1c control less than 9%, less than 8%, and less than 7%, respectively, compared with SC (all P < 0.01) (Table 3). Across all A1c control measures, several factors consistently influenced achievement. Male patients were 48% to 57% more likely than female patients to achieve A1c control (all P ≤ 0.05), and each additional ambulatory visit with the health system was associated with a 14% to 26% greater likelihood of meeting the quality measure (all P ≤ 0.05). Conversely, patients with an index date in the fourth quarter were 69% to 97% less likely to achieve A1c control compared with those seen in the first quarter of the year (all P < 0.01). Higher baseline A1c and greater number of chronic medications at baseline were associated with decreased odds of achieving A1c control across all measures (all P < 0.01). Higher CCI score at baseline was associated with lower odds of achieving A1c control for less than 9% and less than 8% measures (P < 0.05), but not for less than 7%. Patients in the PPCC group were 1.5 times more likely to achieve BP control compared with the SC group (P = 0.04). Detailed fixed effects for all quality measures are presented in Table 3.

TABLE 3.

Multiple Logistic Regression Model for Diabetes Quality Measure Achievement

Effect	OR	95% CI	P value
A1c control <9.0% (C statistic = 0.86)
PPCC vs SC group	3.68	2.31-5.84	<0.01 a
Age (in 1-y increase)	1.03	1.01-1.05	<0.01 a
Male vs female	1.55	1.12-2.13	<0.01 a
Black vs White	1.07	0.76-1.51	0.71
Other vs White	0.86	0.50-1.49	0.59
Smoking status
Current vs never	1.24	0.79-1.93	0.35
Former vs never	1.28	0.89-1.83	0.18
Medicare	1.05	0.65-1.70	0.83
Commercial	1.07	0.73-1.56	0.73
Hypertension diagnosis	0.82	0.55-1.22	0.32
CCI score	0.76	0.59-0.98	0.03 a
Number of ambulatory visits	1.26	1.14-1.40	<0.01 a
Number of ICD diagnosis codes	1.03	0.98-1.08	0.24
Number of chronic medications	0.95	0.91-0.98	<0.01 a
Index date quarter
April to June vs January to March	1.14	0.78-1.68	0.51
July to September vs January to March	0.81	0.51-1.28	0.36
October to December vs January to March	0.23	0.14-0.40	<0.01 a
Baseline values
A1c	0.54	0.48-0.58	<0.01 a
SBP	1.02	1.00-1.03	<0.01
DBP	0.99	0.97-1.01	0.34
A1c control <8.0% (C statistic = 0.81)
PPCC vs SC group	3.53	2.12-5.89	<0.01 a
Age (in 1-y increase)	1.02	1.00-1.03	0.06
Male vs female	1.48	1.09-2.00	0.01 a
Black vs White	0.90	0.65-1.25	0.53
Other vs White	0.59	0.34-1.01	0.05
Smoking status
Current vs never	1.00	0.65-1.53	0.98
Former vs never	1.15	0.82-1.61	0.43
Medicare	1.17	0.73-1.86	0.52
Commercial	0.98	0.67-1.42	0.91
Hypertension diagnosis	1.24	0.84-1.83	0.28
CCI score	0.72	0.56-0.91	<0.01 a
Number of ambulatory visits	1.16	1.07-1.26	<0.01 a
Number of ICD diagnosis codes	1.04	1.00-1.09	0.06
Number of chronic medications	0.92	0.89-0.95	<0.01 a
Index date quarter
April to June vs January to March	1.15	0.80-1.65	0.45
July to September vs January to March	0.85	0.54-1.33	0.48
October to December vs January to March	0.31	0.18-0.53	<0.01 a
Baseline values
A1c	0.60	0.55-0.66	<0.01 a
SBP	1.01	1.00-1.02	0.21
DBP	0.99	0.98-1.01	0.47
A1c control <7.0% (C statistic = 0.76)
PPCC vs SC group	4.61	2.48-8.56	<0.01 a
Age (in 1-y increase)	1.02	1.00-1.05	0.04 a
Male vs female	1.57	1.06-2.33	0.02 a
Black vs White	0.80	0.52-1.21	0.29
Other vs White	0.68	0.33-1.42	0.31
Smoking status
Current vs never	1.03	0.62-1.93	0.35
Former vs never	1.08	0.70-1.67	0.18
Medicare	0.94	0.50-1.77	0.85
Commercial	1.02	0.62-1.68	0.94
Hypertension diagnosis	1.12	0.68-1.86	0.66
CCI score	0.82	0.59-1.14	0.23
Number of ambulatory visits	1.14	1.05-1.23	<0.01 a
Number of ICD diagnosis codes	1.04	0.99-1.10	0.15
Number of chronic medications	0.90	0.86-0.94	<0.01 a
Index date quarter
April to June vs January to March	1.42	0.93-2.18	0.51
July to September vs January to March	0.74	0.41-1.34	0.36
October to December vs January to March	0.03	0.01-0.25	<0.01 a
Baseline values
A1c	0.84	0.76-0.93	<0.01 a
SBP	1.00	0.99-1.02	0.62
DBP	1.00	0.98-1.04	0.92
Poor A1c control >9.0% (C statistic = 0.86)
PPCC vs SC group	0.25	0.15-0.39	<0.01 a
Age (in 1-y increase)	0.97	0.96-0.99	<0.01 a
Male vs female	0.66	0.48-0.92	0.01 a
Black vs White	0.83	0.59-1.18	0.30
Other vs White	1.15	0.66-2.01	0.63
Smoking status
Current vs never	0.86	0.55-1.36	0.54
Former vs never	0.83	0.57-1.20	0.31
Medicare	1.10	0.67-1.80	0.71
Commercial	1.02	0.69-1.49	0.93
Hypertension diagnosis	1.22	0.81-1.83	0.35
CCI score	1.33	1.03-1.72	0.03 a
Number of ambulatory visits	0.78	0.70-0.87	<0.01 a
Number of ICD diagnosis codes	0.97	0.93-1.02	0.24
Number of chronic medications	1.06	1.02-1.09	<0.01 a
Index date quarter
April to June vs January to March	0.90	0.61-1.34	0.60
July to September vs January to March	1.12	0.70-1.80	0.63
October to December vs January to March	4.95	2.90-8.47	<0.01 a
Baseline values
A1c	1.99	1.80-2.20	<0.01 a
SBP	0.98	0.97-0.99	<0.01 a
DBP	1.01	0.99-1.03	0.31
Blood pressure control <140/90 mm Hg (C statistic = 0.76)
PPCC vs SC group	1.49	1.01-2.22	0.04 a
Age (in 1-y increase)	0.99	0.97-1.00	0.11
Male vs female	0.77	0.57-1.03	0.08
Black vs White	0.67	0.49-0.91	0.01 a
Other vs White	1.25	0.71-2.21	0.44
Smoking status
Current vs never	0.87	0.58-1.31	0.49
Former vs never	1.10	0.79-1.54	0.57
Medicare	1.52	0.97-2.39	0.07
Commercial	1.39	0.98-1.99	0.07
Hypertension diagnosis	0.79	0.54-1.14	0.21
CCI score	0.94	0.77-1.22	0.79
Number of ambulatory visits	1.15	1.05-1.25	<0.01 a
Number of ICD diagnosis codes	1.00	0.96-1.04	0.95
Number of chronic medications	0.99	0.96-1.02	0.37
Index date quarter
April to June vs January to March	0.79	0.56-1.13	0.20
July to September vs January to March	1.06	0.68-1.66	0.79
October to December vs January to March	0.57	0.36-0.90	0.02 a
Baseline values
A1c	1.03	0.96-1.11	0.42
SBP	0.95	0.94-0.96	<0.01 a
DBP	1.00	0.98-1.02	0.78

^a P value is statistically significant.

A1c = glycated hemoglobin; CCI = Charlson Comorbidity Index; DBP = diastolic blood pressure; ICD = International Classification of Diseases; OR = odds ratio; PPCC = pharmacist-physician collaborative care; SBP = systolic blood pressure; SC = standard care; y = year.

Discussion

Patients receiving primary care in clinics with a PPCC model were more likely to achieve comprehensive diabetes care quality measures compared with those receiving SC in a community-based health system. Specifically, patients in the PPCC group were 3.7, 3.5, and 4.6 times more likely to achieve A1c control less than 9%, less than 8%, and less than 7%, respectively, had 75% lower odds of poor A1c control (>9%), and were 1.5 times more likely to achieve BP control. Patients seen in the last quarter of the year, having more comorbidities, taking more medications at baseline, and having a higher baseline A1c had lower odds of achieving A1c control across all measures.

The main strength of this study is the methodology used to evaluate the achievement of multiple comprehensive diabetes care quality measures within primary care. The matching process at both the clinic and the patient levels minimized confounding and provided a more accurate assessment of the impact of pharmacists in diabetes care. Propensity score matching ensured that all patients were clinically similar with regard to their demographic characteristics and diabetes control status, addressing limitations of previous studies, such as imbalances in patient characteristics, ^{20 , 26 , 27} single-arm design, ²⁸ or lack of generalizability because of narrowly defined patient cohorts. ¹⁷ This approach, applied within a large community-based health system, offers important insights into the effectiveness of pharmacist integration in routine, real-world clinical practice. ^{33 , 34} Another strength of the study is the larger sample size (total N = 1,263; PPCC group n = 431) compared with previous studies that included between 17 and 207 patients in the pharmacist group, ^{17 , 20 , 26 – 28 , 35 – 37} with the exception of one. ³⁸

Our finding that pharmacists had a significant positive impact on A1c control aligns with previous literature. ^{19 , 20 , 23 , 28 , 35 , 38 – 41} Unlike other studies that evaluated HEDIS quality measures using A1c cut-off points of less than 7% or less than 9% at various time points throughout the year, ^{26 , 29} our study adheres to the original intent of these measures by assessing A1c control (less than 7%, 8%, and 9%) at the end of the calendar year, while controlling for the quarter of the year in which the patient had their first clinic visit. ³ Because it more closely aligns with real-world clinical and policy contexts, this methodological difference enhances the applicability of our results to payers and health systems.

The influence of temporal factors on quality measure achievement, as highlighted in our study, raises important questions about the appropriateness of static, annual cut-offs for quality measures. Patients who were seen in clinics in the fourth quarter of the year were 67% less likely to achieve A1c control of less than 9%, with similarly substantial reductions in likelihood for targets of less than 8% (60%) and less than 7% (95%). These findings underscore an important limitation of current quality measures in that they do not account for the timing of interventions, potentially penalizing health care professionals who initiate treatment later in the calendar year when there may be insufficient time for improvements to manifest in year-end assessments. By identifying this temporal aspect, our study reveals a critical gap in current quality measure assessment for diabetes care, emphasizing the need for more dynamic evaluation approaches.

In our analysis, patients in the PPCC group had fewer ambulatory visits than those in the SC group (2.3 vs 3.5) but were significantly more likely to achieve the A1c control quality measures. This contrasts with a previous study examining the effect of a similar PPCC model on time to BP goal, where patients in the PPCC group experienced more frequent encounters with the pharmacist and lower therapeutic inertia. ⁴² One possible explanation is that the clinical encounters in the PPCC group were more targeted to diabetes management compared with SC, resulting in improved A1c control. Although not assessed in our study, pharmacists at BSMH are accustomed to performing treatment intensification to ensure that the patients achieve optimal diabetes outcomes without the need for additional visits.

A1c control was inversely associated with both the number of chronic medications and the CCI score, which is consistent with existing literature suggesting that patients experiencing polypharmacy or those with a higher number of comorbidities face greater challenges in achieving or maintaining their clinical goals. ^{43 – 46} Particularly, polypharmacy is frequently associated with medication nonadherence, which can hamper effective A1c control. ⁴⁷ Male patients were more likely to achieve A1c control than female patients, which also aligns with previous research. ^{48 – 50}

Finally, patients with higher baseline SBP were less likely to meet the quality measures A1c greater than 9 and BP less than 140/90 by the end of the calendar year. This pattern is consistent with the observations for baseline A1c values, in both this and previous studies. ^{51 , 52} These findings underline that health status at the index visit significantly affects end-of-year quality measure achievement. Given this relationship, the earlier the pharmacist intervention in diabetes care, the better. Pharmacist involvement could facilitate timely interventions to promote both A1c and BP control in patients with T2DM, resulting in improved quality measure achievement.

Our study has several implications for practice and policy. Health systems should consider integrating pharmacists into primary care teams to improve the quality of diabetes care and, consequently, increase reimbursement through quality measure achievement. Additional sources of revenue to justify a pharmacist salary could be through offering Medicare annual wellness visits or chronic care management services. In 2019, a private family medicine clinic generated revenue as high as $72,534.12 for pharmacist-led Medicare annual wellness visits and $5,922.47 for chronic care management. ⁵³ Furthermore, several strategies to reduce wait times for primary care appointments have been identified in the literature. One such strategy consists of using other health care professionals, including pharmacists, to increase the team’s capacity to address specific patient needs. This allows the primary care physician to provide more effective patient care and reduce wait times for primary care visits. ⁵⁴ A recent study estimated a reduction of 640 hours per year in the primary care physicians’ workload when pharmacists manage chronic conditions under a collaborative practice agreement. ⁵⁵ Proposed full-time equivalents (FTEs) for pharmacists in primary care range from 0.25 to 1.0 FTE, depending on the risk assessment of the patient panel. ⁵⁶ One FTE pharmacist can open up 1,920 primary care physician appointments per year. ⁵⁵ Finally, studies examining the economic impact of pharmacist-led diabetes management in the United States have shown a decrease in overall health care spending compared with usual care, ⁵⁷ cost-effective interventions over extended time horizons, ⁵⁸ and greater cost avoidance with appreciable return on investment, including lower cost per quality-adjusted life-year gained compared with usual care. ⁵⁹

Our findings also highlight limitations in current quality measures in diabetes care. These measures fail to reflect the complexity of diabetes management and to capture individual patient progress. ⁶⁰ For instance, a patient reducing their A1c from 12% to 9.5% constitutes a substantial clinical improvement, yet it would still be considered noncompliant with the A1c less than 9% quality measure. The Agency for Healthcare Research and Quality emphasized the need for more comprehensive measures. ⁵¹ Future research should focus on developing quality measures that consider patient heterogeneity, baseline values, individual trajectories, and timing of care initiation.

LIMITATIONS

This study is not without limitations. First, even though the study was multisite, the EHR data analyzed pertained to a single health system; thus, the information available relates to services delivered within the BSMH network only, which may not provide a comprehensive view of the patients’ health status. Further, we did not have access to prescription refill data and therefore could not confirm our determination of what constituted chronic medication use. Second, the observational and retrospective nature of the study limits our ability to assess causality. Third, although we matched clinics on various characteristics, we did not have data on clinical performance or outcomes during the design or analysis phase. Thus, our findings could be positively biased if the matching resulted in the comparator clinics having poorer outcomes than the PPCC group. Additionally, despite the use of propensity score matching and generalized linear mixed models, the possibility of unmeasured confounding variables cannot be ruled out entirely. Fourth, although we controlled for total ambulatory visits, we could not assess the specific nature and focus of these visits, so the higher number of visits verified in the SC group may have been for reasons other than diabetes management. Finally, there was a lack of standardization in the pharmacist intervention. Although the primary criterion for physician referral to a pharmacist in the PPCC group was an A1c greater than or equal to 9%, referrals also occurred at lower A1c levels (≥7%). Additionally, not all eligible patients were referred to or followed by a pharmacist, potentially introducing selection bias. Despite these limitations, the study’s large sample size, rigorous statistical methods, and real-world setting provide valuable insights into PPCC in diabetes management. The use of propensity score matching and data from multiple clinics strengthens the generalizability of our findings, providing robust evidence for the effectiveness of pharmacist-led diabetes management in primary care.

Conclusions

Patients with uncontrolled diabetes who received primary care in clinics with a PPCC model were more likely to achieve comprehensive diabetes care quality measures related to A1c and BP control than those receiving SC. These findings demonstrate the effectiveness of integrating pharmacists into primary care teams for diabetes management, providing a strong foundation for health care systems to consider implementing similar models to increase diabetes care quality. This approach aligns with the ongoing shift toward team-based and value-based care in an effort to enhance the overall quality of care delivered.