The Impact of Primary Care Practitioner Intervention as an Adjunct to a Diabetes Case Management System

Abstract

Objective

The objective of this study was to assess whether additional primary care practitioner (PCP) contacts beyond the intake visit are associated with reduced hemoglobin A1c in patients with type 2 diabetes actively engaged in the Kaiser Permanente case management system.

Methods

This retrospective cohort study using the Kaiser Permanente electronic health record explored the effect of enhanced PCP contact among adult patients with type 2 diabetes actively working with diabetes case managers (defined as ≥ 4 case manager contacts during the study period).

Results

A total of 837 patients met the inclusion and exclusion criteria. On average, patients with the highest PCP contact, < 7 contacts, had Ac levels 0.53 lower than those in the lowest PCP contact quartile, < 3 contacts (p = 0.0007). A1c decreased an average of 0.20 when the PCP contact quartile was one quartile higher (p = 0.0004). Holding the baseline A1c constant, the A1c decreased an average of 0.15 when the PCP contact quartile was one quartile higher (p = 0.0024). A1c change was significantly correlated with baseline A1c; A1c decreased by 0.64 more as the baseline A1c level increased by 1 (p < 0.0001). Additionally, the A1c level decreased by 0.02 more when patient age increased by 1 (p < 0.0001). Metformin use was associated with a decrease of A1c by 0.40 (p = 0.0057), whereas insulin use was associated with an increase of A1c by 0.29 (p = 0.0280).

Conclusion

In summary, a significant reduction was observed in A1c in patients with increased PCP contacts. This effect was seen in patients already receiving recommended case manager support.

Introduction

Medical therapy to decrease glucose levels has been shown to be associated with long-term microvascular and macrovascular complications caused by uncontrolled diabetes.¹ However, achieving glycemic control has been traditionally difficult and requires a cooperative and multidisciplinary approach. Optimal treatment plans require patient self-management with medication and adherence to a carbohydrate-controlled diet, in addition to regular exercise. Patients with poor glycemic control (such as hyperglycemia and elevated A1c) may also reflect insufficient and inadequate practitioner follow-up and intervention strategies.² Addressing these barriers can be time-consuming and represents a formidable challenge to overcome for the already taxed primary care practitioners (PCPs). Past studies have demonstrated that nurse case managers can play an important role in improving important diabetes outcomes such as improving A1C levels, blood pressure, lipids, and depression scores.^3–7 Improvement was most pronounced when case managers with a large amount of diabetes experience were able to make management changes without having to wait for PCPs to review their choices, presumably due to avoiding delays in care changes and overcoming care inertia.

Kaiser Permanente in Southern California has implemented a similar treatment team approach utilizing nurse case managers who work with patients with uncontrolled diabetes (A1c greater than 8). Data shows that the number of successful interventions by the case manager improves glycemic control (defined as improved A1c) and that certain populations require a larger number of “touches” before an effect can be demonstrated (such as non-English speakers).⁸ After an initial visit with a primary care doctor, patients are followed closely by case managers, and medications are adjusted based on PCP-approved algorithms. These case managers may also schedule visits with diabetes educators, make laboratory order, and advise primary care doctor follow-up to move care forward or for management that falls outside their algorithms. This close involvement is successful in reducing A1c values.⁸ Previous studies and internal Kaiser Permanente data have shown the efficacy of case management with at least 4 case manager interventions per year.^3–5,8

However, the specific impact of the PCP on glycemic control after the initial visit and the amount of involvement necessary to provide benefit have not been well characterized. Studies in health maintenance screening have demonstrated that a key barrier to changing patient behavior is the lack of recommendations from a PCP.⁹ Practitioners thus have an influential role in stimulating patient behavior changes and managing diabetes in the context of other comorbidities. It is thus important to determine the optimal role and level of involvement of the PCP in an established case management system.

Methods

This is a retrospective cohort study where the Kaiser Permanente San Bernardino County database of electronic health records was used to search for patients above the age of 18 with diagnosed type 2 diabetes mellitus with at least 4 case manager contacts between October 9, 2017, to October 9, 2018. Patients needed to have continuous Kaiser Permanente membership (allowing gaps that are less than 45 days). Pregnancy, end stage renal disease, liver disease, anemia, thalassemia, hemoglobinopathies, or therapies that may affect A1c values (Epogen, blood transfusion, or steroid therapy) were among the exclusion criteria. Comorbidities and exclusionary diagnoses were identified using International Classification of Diseases (ICD) codes (see Figure 1). Medication use was ascertained from pharmacy data. Baseline A1c was assessed in a 3-month window (between 45 days prior and 45 days after the trial start date of October 9, 2017). PCP was determined from the electronic medical record. Diabetes-related patient encounters (telephone encounters, office visits, email encounters) were identified if there was a medication change or laboratory order associated with diabetes ICD codes (ICD-9 25-, ICD-10 E11, or E13). The A1c at the trial end, October 9, 2018, was assessed in a similar 3-month window (between 45 days prior and 45 days after October 9. 2018). Approval from the institutional review board was obtained.

Figure 1:

Cohort Selection. ESRD = End Stage Renal Disease; ICD = International Classification of Diseases.

The primary endpoint was the percent and absolute change in A1c compared to baseline. Linear regression was used to explore the association between A1c change (dependent variable) and the number of PCP interventions (independent variable). The data was also analyzed to look for differences in the association based on covariates of age, gender, race/ethnicity, medical comorbidities, medications used, body mass index (BMI), baseline A1c, and the number of case manager contacts. Analysis was performed both as continuous variables as well as categorical variables based on the quartiles defined by the distribution of intervention counts (PCP contacts < 3, = 3, 4–6, ≥ 7). Comparisons were analyzed with the Kruskal-Wallis test for continuous variables and chi-square or Fisher’s exact test for categorical variables. A 2-sided p value < 0.05 was considered statistically significant. All statistical analysis was performed using SAS 9.4 (SAS Institute Inc., Cary, NC).

Results

A total of 837 adult patients who met all the inclusion and exclusion criteria were identified with type 2 diabetes mellitus. Baseline clinical characteristics of the control and treatment groups are shown in Table 1. The average age of the cohort was 56.3 ± 10.63 years, and their average BMI was 34.04 ± 7.05. Of the study participants, 45.88% were female and 54.12% were male. Most of the patients were Hispanic (59.38%), followed by non-Hispanic White (21.39%), non-Hispanic Black (9.56%), Asian and Pacific Islander (7.05%), and other or unknown (2.63%). Table 2 contains descriptive statistics for the trial population categorized into quartiles. The mean total number of case manager contacts was 11.54 ± 7.83, with a lower quartile of 6 and upper quartile of 14. Mean PCP contacts was 5.03 ± 3.45, with a lower quartile of < 3 and upper quartile of 7. Specifically, 198 patients were contacted by the PCP < 3 times, 125 patients were contacted 3 times, 295 patients were contacted 4–6 times, and 219 patients were contacted 7 or more times. The A1c at baseline was found to be 9.43 ± 1.46%, with a baseline lower quartile of 8.4% and upper quartile of 10%. The minimum baseline A1c of trial participants was 7.1%, with a maximum of 18.1%.

Table 1:

Patient demographics, comorbidities, and medications used

Demographics	Number of patients	Percentage (%)
Sex
Female	384	45.88
Male	453	54.12
Race/Ethnicity
Non-Hispanic White	179	21.39
Non-Hispanic Black	80	9.56
Hispanic	497	59.38
Asian and Pacific Islander	59	7.05
Other/Unknown	22	2.63
Comorbidities
Coronary artery disease	81	9.68
Peripheral vascular disease	31	3.70
Stroke	15	1.79
Diabetic nephropathy	237	28.32
Diabetic retinopathy	215	25.69
Diabetic neuropathy	353	42.17
Diabetic foot ulcer	32	3.82
Dyslipidemia	710	84.83
Medications used
Biguanides	672	80.29
Sulfonylureas	392	46.83
Biguanides and sulfonylureas	12	1.43
Thiazolidinediones	35	4.18
Alpha-glucosidase inhibitors	4	0.48
DPP-4 inhibitors	35	4.18
SGLT-2 inhibitors	8	0.96
GLP-1 Agonists	9	1.08
Insulin	561	67.03

DPP-4, Dipeptidyl peptidase 4; GLP-1, glucagon-like peptide 1; SGLT-2, Sodium-glucose Cotransporter-2.

Table 2:

Descriptive statistics and hemoglobin A1c change organized by the number of primary care physician interventions and categorized into quartiles

Variables	Mean ± SD	Lower quartile	Median	Upper quartile	Min	Max
Age (in years)	56.3 ± 10.63	49.87	57.44	64.03	19.23	74.73
BMI	34.04 ± 7.05	29.07	32.88	38.14	19.65	60.55
Case Manager contacts	11.54 ± 7.83	6	9	14	4	57
PCP contacts	5.03 ± 3.45	3	4	7	0	24
A1c baseline (%)	9.43 ± 1.46	8.4	9	10	7.1	18.1
A1c outcome (%)	8.92 ± 1.64	7.7	8.7	9.9	5.4	15.9
A1c change (%)	−0.51 ± 1.80	−1.3	−0.4	0.5	−10	6.7

BMI, body mass index; PCP, primary care practitioner; SD, standard deviation.

Linear regression analysis demonstrated that, on average, the A1c level increased by 0.53 when comparing the lowest PCP contacts quartile < 3 with the highest quartile ≥ 7 (p = 0.0007). On average, A1c decreased by 0.20 when the PCP contact quartile was one quartile higher (p = 0.0004). Holding the baseline A1c constant, A1c decreased an average of 0.15 when the PCP contact quartile was one quartile higher (p = 0.0024).

A1c change was significantly correlated with baseline A1c. A1c decreased by 0.64 more when the baseline A1c level increased by 1 (p < 0.0001). Additionally, holding other variables constant, the A1c level decreased by 0.02 more when patient age increased by 1 (p < 0.0001). Metformin use was associated with a decrease of A1c by 0.40 (p = 0.0057), whereas insulin use was associated with an increase of A1c by 0.29 (p = 0.0280). There was no statistically significant correlation with the use of other antihyperglycemic medications or medical comorbidities (listed in Table 1). There was also no statistically significant correlation between A1c change, BMI, or the number of case manager contacts of the trial participants. In terms of the other categorical variables, there was no statistically significant association between A1c change and gender or with race/ethnicity.

Discussion

Given the chronic nature and preventable complications of type 2 diabetes mellitus, gaps in patient care can be detrimental to patient health outcomes.¹⁰ To improve diabetic outcomes, an elaborate interplay at the patient, practitioner, and system level is required, all while recognizing the various socioeconomic and cultural barriers to diabetes management.¹⁰ On the practitioner level, there has been evidence that nurse case manager support (≥ 4 “touches” per year) has positively affected important diabetic outcomes. The authors thus studied a population of patients with at least this level of case manager support to quantify the additive benefits of PCP contacts in this cohort. These findings, with the primary endpoint being A1c levels, demonstrate that additional PCP contacts can aid in the reduction of A1c. Improvement in A1c was seen as soon as PCP contacts reached 3 contacts within a year, with additional improvement seen as the number of contacts increased. This effect is seen in addition to at least 4 case manager contacts. Significant decreases in A1c were also reported in individuals with greater ages and baseline A1c values.

Metformin and insulin also seemed to affect the change in A1c. Metformin use was associated with A1c improvement, but insulin therapy was unexpectedly associated with worse A1c. This discovery is most likely the result of a combination of factors. A patient who is placed on insulin may theoretically have more difficult to control diabetes and as such is more likely to have worsening A1c compared to a patient who ultimately does not require insulin. For example, A1c may worsen in a patient whose reasons for poor control cannot be entirely compensated for with insulin (such as poor lifestyle [diet, exercise] choices), especially as insulin is initiated and the therapeutic dose has not yet been identified. Insulin adherence may also be poorer than non-insulin medicine adherence.¹¹ Furthermore, this may suggest a limitation of the case management system, and other measures may need to be employed for insulin-requiring patients to have better success. These findings might indicate that extra PCP assistance and/or other assistance is needed in specific subpopulations.

A strength of this study is the application of a large electronic medical record database to assess the diabetes care of the patients in this study. As members of a managed care organization, most Kaiser Permanente patients receive their care entirely within the Kaiser Permanente network, allowing for a more comprehensive database.

Limitations

However, this study has several limitations, aside from the inherent biases associated with retrospective comparisons. First, it is possible that not all PCP contacts were captured, as some practitioners may not have documented utilizing the prespecified ICD codes in some of their diabetes-related patient contact. The authors also focused only on encounters where a medication change or laboratory order was associated to a diabetes diagnosis code; this study focused on those encounters because they were more likely to have involved a specific discussion of diabetes. It is possible that the authors did not capture additional diabetes discussions that did not require a medication or laboratory order. Secondly, patients who obtain regular A1c levels may be more adherent with medications and lifestyle modifications or may have resources, support systems, or other things that allow them to be engaged in their diabetes care. Furthermore, patients followed more closely by PCPs may be doing so primarily for reasons unrelated to diabetes (for example, management of other comorbidities). Third, data were not collected on the specific medication changes that were made during the study period. Lastly, this study was limited to San Bernardino County Kaiser Permanente members and may not be generalizable to other regions or health plans.

Conclusion

More PCP contacts appear to be associated with better glycemic control (improved A1c values) even in a setting of enhanced case management. With benefits seen with as few as 3 interventions per year, PCP contact at least every 4 months may be one strategy that would incorporate the findings of this study. PCP intervention should also be done while maintaining communication with the extended diabetes team (including case managers) to provide the patient with a cohesive plan. Further studies are warranted to further characterize the optimal system utilizing both case managers and physicians, as well as the nature and types of intervention that are most high yield or which may set the PCP intervention apart from case manager intervention. It would be helpful to explore whether interventions outside of medications or laboratory orders are helpful, including those related to mental health, housing/transportation support, and the expense/cost of medical care. Future studies may also be aimed at the effect of additional PCP involvement regarding other clinically important diabetic metrics (like blood pressure, lipid levels, and depression scores), as well as the impact of additional factors related to diabetes history/care (including time since diagnosis, A1c trend prior to intervention, and frequency of medication changes).