Evaluation of four disease management programs: evidence from blue cross blue shield of Louisiana

Abstract

Aims: Chronic diseases impose a substantial healthcare burden. This study sought to evaluate the clinical and economic impact of new disease management (DM) programs, targeting four major chronic disease groups: diabetes, coronary heart disease (CHD)/hypertension (HTN), asthma/chronic obstructive pulmonary disease (COPD), and congestive heart failure (CHF)/chronic kidney disease (CKD).

Materials and methods: Between March 1, 2015, and February 28, 2018, members with Blue Cross Blue Shield of Louisiana insurance were contacted and enrolled in a DM program if they were aged 18 years through 64 years, eligible for a DM program, and had not been previously enrolled in a DM program. Active enrollees of a DM program (“IN” group) were compared to members who were not yet enrolled (“OUT” group). Average per member per month (PMPM) costs were aggregated annually to document any descriptive trends. Multivariable model estimates were used to compare PMPM costs for all IN subjects and all OUT subjects. Total medical savings were evaluated for the following time intervals: 1–12 months, 13–24 months, and 25–36 months.

Results: For all four DM programs, average costs PMPM trended upward over time for the OUT cohort, while they remained relatively stable for the IN cohort. Some evidence also showed that DM programs improved clinical outcomes, such as hemoglobin A1c values. A difference in difference analysis showed PMPM savings for all four programs combined of $31.61, $50.45, and $53.72 after 1, 2, and 3 years, respectively. Multivariable modeling results showed total savings after 3 years of $14,460,174 for all DM programs combined.

Limitations: Although multivariable models adjusted for several clinical, demographic, and economic characteristics; it is possible that some important confounders were missing due to lack of data.

Conclusions: DM programs implemented to control diabetes, CHD/HTN, CHF/CKD, and asthma/COPD are cost-effective and show some evidence of improved clinical outcomes.

Keywords:

• Disease management
• chronic disease
• healthcare expenditures
• nurse intervention
• clinical outcomes

JEL CLASSIFICATION CODES:

• C31
• I13
• M21

Introduction

Chronic disease burden

Chronic conditions are responsible for a substantial clinical and economic healthcare burden, accounting for 70% of deaths and 86% of healthcare costs in the United States¹. In 2014, about 60% of Americans had at least 1 chronic disease (18% had 1 chronic disease, 42% had 2 or more), and that percentage is expected to increase as the population ages and grows^2,3. Asthma, chronic obstructive pulmonary disease (COPD), diabetes, hypertension (HTN), coronary heart disease (CHD), chronic kidney disease (CKD), and congestive heart failure (CHF) are among the costliest chronic disease groups in the United States and affect millions of Americans each year^4,5. However, the burden of chronic illness is more substantial in some states compared to others.

For example, Louisiana has the second-highest diabetes mortality rate in the nation, with approximately 13.9% of the adult population having diabetes⁶. Louisiana also ranked second behind Mississippi in age-adjusted kidney disease mortality rates per 100,000 people in 2017⁷. Additionally, the age-adjusted prevalence of self-reported HTN among adults in Louisiana was 37.5% in 2015, whereas the national average was 29.8%⁸. Data from the Behavioral Risk Factor Surveillance System between 2014 and 2015 showed that the age-adjusted prevalence of self-reported, physician-diagnosed COPD among adults in Louisiana was about 7.1%; and, 45.3% of those patients listed asthma as a comorbidity⁹. Furthermore, Louisiana has the sixth-highest adult obesity rate in the nation at 36.2% (up from 22.6% in 2000 and 12.3% in 1990); and, while it is not always considered a chronic condition itself, obesity is associated with a high risk for other chronic illnesses like diabetes, HTN, and cardiovascular disease^10–12. As of 2015, an estimated 2.9 million people in Louisiana had at least 1 chronic illness; and, 1.2 million had 2 or more¹³.

Chronic conditions (diabetes, CHD/HTN, asthma/COPD, and CHF/CKD) have serious clinical implications, but they are also expensive to patients, providers, and payers alike. The high prevalence of chronic illness in Louisiana is an indication that the economic burden attributable to chronic disease in Louisiana is also substantial. According to the Partnership to Fight Chronic Disease¹³, the aggregated healthcare expenditures attributed to chronic conditions in Louisiana between 2016 and 2030 are projected to total $612 billion. They also report that chronic illness in Louisiana could cost $28.8 billion in medical costs alone.

Despite the high prevalence and sizeable clinical and economic burden of chronic disease in Louisiana and the United States, early diagnosis, lifestyle changes, and treatment can improve quality of life as well as survival^14–18. One way to implement treatments and lifestyle changes that mitigate the effects of chronic illness and improve quality of life is through disease management (DM) programs.

Chronic disease management programs

DM programs offer the potential to prevent clinical events and reduce healthcare expenditures associated with chronic conditions, but available literature regarding the efficacy of DM programs remains inconclusive. Systematic reviews of DM efforts report conflicting evidence^19–21. One such review by de Bruin et al. examined the impact of DM programs on healthcare expenditures for patients with diabetes, depression, heart failure, or COPD²². The review assessed 21 DM studies, each of which reported the incremental healthcare costs of DM programs per patient per year. De Bruin and colleagues found that only 13 of those 21 studies (61.9%) showed cost-savings, with incremental costs ranging between −$16,996 and $3,305 per patient per year. Results likely varied so widely because the researchers reported large discrepancies across studies in terms of design, the components of DM programs, interventions within components, and characteristics of economic evaluations²². Nevertheless, DM programs are increasingly implemented in healthcare systems worldwide; and, more recent systematic reviews recommend new approaches, such as reorganizing clinical practice and allowing non-physician practitioners to adjust prescription medication therapy²³.

Given the significant clinical and economic burden that chronic illness places on Louisiana, Blue Cross Blue Shield of Louisiana (BCBSLA) constructed a DM program to help improve medical outcomes, reduce healthcare costs, and better control chronic diseases. The DM program was designed to target patients with chronic diseases and high-risk behaviors (smoking, poor nutrition, inactivity, etc.) and focuses on the philosophy that increased participation and self-management by members who are chronically ill will result in improved overall health. The program also emphasizes using a nurse care-manager or health coach in conjunction with social workers, dietitians, pharmacists, and clinicians. Initially, the program was only available to certain group members and some conditions; however, in 2014, BCBSLA expanded the DM program to all members with a diagnosis of at least 1 chronic condition to better serve the needs of its population. The present study aims to evaluate the DM program at BCBSLA, now that it is more widely available, for 4 major groups of chronic diseases: (1) diabetes, (2) CHD/HTN, (3) asthma/COPD, and (4) CHF/CKD.

Methods

BCBSLA DM program referral and enrollment process

BCBSLA members who meet the eligibility criteria for a particular DM program are referred to and automatically enrolled in that program. Eligibility requirements vary based on the member’s chronic illness (described in Table 1); and, referrals are triggered by a member’s healthcare utilization via the monitoring and analysis of claims data (e.g. emergency department [ED] visits, inpatient stays, outpatient visits, prescription fills, etc.). Patient referral lists are generated monthly and depend on a hierarchy of conditions for members with more than 1 chronic illness. For example, if a member has both asthma and diabetes, they would only be referred to the diabetes DM program but would receive guidance and management strategies for both conditions through the diabetes DM program. The hierarchy of conditions for this study (ranked lowest to highest, also depicted in Table 1) is as follows: asthma, CHD/HTN, COPD, diabetes, CHF, and CKD.

Table 1. Eligibility requirements for DM Program referral by chronic condition.

Abbreviations. DM, disease management; ED, emergency department; IP, inpatient; OP, outpatient; HTN, hypertension; CHD, coronary heart disease; AMI, acute myocardial infarction; CABG, coronary artery bypass grafting; PCI, percutaneous coronary intervention; COPD, chronic obstructive pulmonary disease; CHF, congestive heart failure; CKD, chronic kidney disease.

* Professional visits exclude visits for radiology or lab work.

After identifying eligible patients with a chronic disease of interest through claims analysis, BCBSLA refers qualifying members to a given DM program, thereby enrolling the member. New enrollees receive a welcome packet in the mail and an introductory call from a health coach. Members who are not contacted successfully are unenrolled in the program after 45 days, or they may opt-out of the program at any time. Those members who are contacted successfully may transfer to a nurse immediately during business hours or request a call back from a medical professional after business hours. High-risk members receive their introductory call from a nurse directly, although all DM program enrollees have access to a nurse via phone.

Members who are engaged in the DM program receive guidance from a health coach and a multidisciplinary team of healthcare professionals at BCBSLA, which includes physician advisors, registered nurses, licensed practical nurses, pharmacists, dieticians, and behavior health specialists. This multidisciplinary team participates in case conferences to coordinate care for individuals who require more intensive clinical management. Nurses also work with a variety of providers to assist patients with obtaining healthcare services. All DM program nurses are trained in motivational interviewing and coaching techniques; and, after they determine the acuity of a patient’s condition, they recommend the frequency of scheduled calls that the DM program participant should have with their health coach (patients with severe conditions may have regularly scheduled calls directly with the nurse at more frequent intervals). Participants are expected to speak with their health coach at the predetermined intervals to be considered actively enrolled in the DM program.

Inclusion criteria and cohort definitions

This retrospective study utilized anonymized data from BCBSLA from 1 March 2015, through 28 February 2018, to evaluate members who would eventually become actively enrolled in a DM program. Accordingly, all members were required to be actively enrolled in 1 of the 4 DM programs of interest by February of 2018 to be included for analysis. Members were defined as being actively enrolled if they consistently interacted with health coaches and nurses over the study period at their regularly scheduled intervals (higher acuity patients may have frequent calls).

After becoming actively enrolled, members were also required to remain continuously active, or engaged, month to month. Members who remained actively enrolled in a DM program from the day of engagement through the end of the study period were considered continuously engaged (i.e. no drop-out months once actively enrolled). If a patient dropped out for any reason, they were excluded because they did not meet the continuously engaged criterium.

Additionally, all members had to be new to the DM program because we only wanted to evaluate the effects of the DM program after its expansion to more disease categories and group types; thus, anyone who had participated in a previous iteration of the BCBSLA DM program (before 1 March 2015) was excluded from analysis. Patients who were newly enrolled in DM programs for end-stage renal disease (ESRD), prediabetes, or rare conditions were excluded from the study. Furthermore, all members also had to be aged between 18 years and 64 years at the time of DM program referral to be included for analysis.

Because the expanded DM program was offered to all members with the chronic conditions of interest, we could not compare a standard treatment and the control group. Therefore, we utilized the timing of when a member became actively enrolled in a DM program to create a comparator cohort (i.e. comparing those who had already become actively enrolled to those who had not yet become actively enrolled). Throughout the study period, a member was considered part of the “IN” group the month that they actively enrolled in a DM program; and, they remained in the “IN” group until the last month of year 3 (continuously active). Before active enrollment, that member would have been counted in the “OUT” group. Members were considered part of the “OUT” group if they received the introductory call but chose not to call back or speak directly with a nurse.

For example, a patient was eligible for a DM program and referred for enrollment in month 3 of the study period. However, they did not respond to the call and were considered “OUT”. Their condition worsened, so, they decided to become actively enrolled in a DM program in month 22 and remained continuously engaged in the DM program through month 36. This member would be counted as part of the OUT group for months 3–21 and would be part of the IN group for months 22–36. Members essentially served as their own control, allowing us to compare them to themselves as well as other study participants. This method helped with sample size considerations because, by month 12 of year 3, all members were actively enrolled in a DM program.

Members who declined enrollment or were unreachable and never became actively enrolled were excluded from analysis. Similarly, passively enrolled members, who were defined as members that enrolled in the DM program but did not actively engage with their health coach or nurses, were excluded. Passively enrolled members were not included in our analysis because most of these patients were lower acuity patients who felt they could manage on their own, and it is impossible to tell whether they read the educational DM program materials that were sent to them. They could be actively engaged in other ways or not at all, so for this initial study, we chose to focus on people who were willing and actively seeking help through the BCBSLA DM program versus those who concretely refused or ignored help through the BCBSLA DM program to evaluate program effectiveness.

Study design

A stepped-wedge design similar to that used in randomized trials was employed for this retrospective study, in which all subjects were not enrolled in the program initially but were gradually enrolled over time until by the end of the third year, all subjects were enrolled in 1 of the 4 separate DM programs for the following conditions: diabetes, CHD/HTN, asthma/COPD, and CHF/CKD. This analytical methodology allowed us to systematically evaluate the DM programs and maximize participant inclusion due to rolling program enrollment and engagement status. Average per member per month (PMPM) costs were aggregated each year to account for the decreasing sample size associated with the stepped-wedge study design. PMPM data collected for each patient within a 1-month wash-out period before and after initial DM engagement were excluded from the analysis to reduce regression to mean. Select clinical outcomes (hemoglobin A1c [HbA1c] in the diabetes program and systolic and diastolic blood pressure [BP] in the CHD/HTN program) were also evaluated and aggregated annually.

Statistical analysis

Multivariable model estimates were used to compare PMPM costs for all IN subjects and all OUT subjects. It was not feasible to estimate these models separately for each DM program due to sample size considerations and because patients with multiple conditions were only enrolled and managed through 1 of the programs (based on the hierarchy of conditions presented in Table 1). Total medical savings were evaluated monthly and aggregated for the following time intervals: 1-12 months (1 March 2015, through 29 February 2016), 13–24 months (1 March 2016, through 28 February 2017), and 25–36 months (1 March 2017, through 28 February 2018). PMPM costs were evaluated monthly based on enrollment status because of the stepped wedge design. Select patient characteristics (age and sex) for each group over the study period were recorded and are presented in the results.

Savings from the DM programs from a payer perspective were estimated using a difference-in-difference (DiD) approach of trends calculated through a generalized linear regression mixed model²⁴. Essentially, we compared the trends of 2 slopes calculated through the model: (1) the IN group versus (2) the OUT group. The DiD of the IN versus the OUT gave us our savings aggregated yearly for all DM programs combined. The data were analyzed using the PROC GLIMMIX procedure in SAS@ software²⁵. PROC GLIMMIX is a procedure for fitting generalized linear mixed models. It allows for non-normal data, random effects, and correlation amongst responses. The analysis looked at the immediate intervention effect and the effect of time on intervention efficacy for the main outcome variable (e.g. PMPM costs) and used the following basic equation: (1) $$Y=\mathrm{ }{\beta }_0+{\beta }_{t}T+{\beta }_{i}I+{\beta }_a{T}_a\mathrm{ }$$(1) where Y is the main outcome; β₀ is the model intercept; β_t is the model parameter representing the pre-intervention slope; T is the time from the start of the observation period (in months); β_i is the model parameter representing the immediate intervention effect, additive to the intercept; I is a binary variable for intervention status (0 before intervention, 1 after intervention); β_a is the model parameter representing the change in pre-intervention slope after the intervention and is interpreted as a gradual change to the time series following the intervention (the value of the post-intervention slope is [β_t + β_a]); and T_a is the time after the start of the intervention (in months). Again, model results represent medical PMPM trends and DiDs for all DM programs combined due to sample size considerations.

Results

Of all patients enrolled in 1 of the 4 DM programs of interest between 1 March 2015, and 28 February 2018, a total of 8,132 members were aged between 18 years and 64 years at the time of enrollment, new to the DM program, and actively enrolled within the study period. During that time window, however, 2,541 (31%) members were excluded because they had at least 1 drop-out month, leaving 5,591 (69%) members who were actively enrolled and continuously engaged in 1 of the 4 DM programs of interest by February of 2018. The breakdown of participants by DM program is as follows: 3,807 patients were enrolled in the diabetes program, 1,112 patients in the CHD/HTN program, 493 patients in the asthma/COPD program, and 179 patients were enrolled in the CHF/CKD program. Table 2 shows select patient demographic information (age and sex) for each DM group through month 11 of year 3. The total number of patients varied year to year because members could become eligible on a rolling basis, and members could be counted twice in the year they switched from the “OUT” group to the “IN” group. Unadjusted trends for IN and OUT patients are discussed for individual DM programs first, followed by the results from the multivariable analysis for all 4 DM programs combined.

Table 2. Patient demographics by year for the IN cohort versus the OUT cohort^a.

Demographics by DM Program, no. (%)^b	Year 1		Year 2		Year 3^c
	IN	OUT	IN	OUT	IN	OUT^c
Asthma/COPD
Total N	201 (100)	484 (100)	452 (100)	97 (100)	493 (100)	26 (100)
Age, year, mean (SD)	45.0 (0.63)	46.6 (0.29)	45.6 (0.28)	51.6 (0.51)	46.1 (0.25)	53.7 (0.96)
Female sex	136 (67.7)	329 (68.0)	299 (66.2)	69 (71.1)	334 (67.7)	23 (88.5)
CHD/HTN
Total N	394 (100)	1085 (100)	864 (100)	524 (100)	1112 (100)	161 (100)
Age, year, mean (SD)	57.4 (0.15)	56.8 (0.07)	57.4 (0.07)	56.3 (0.11)	57.0 (0.06)	56.9 (0.27)
Female sex	235 (59.6)	671 (61.8)	528 (61.1)	332 (63.4)	687 (61.8)	99 (61.5)
CHF/CKD
Total N	73 (100)	171 (100)	140 (100)	68 (100)	179 (100)	24 (100)
Age, year, mean (SD)	57.8 (0.31)	58.1 (0.15)	58.3 (0.15)	57.8 (0.24)	58.1 (0.14)	58.5 (0.49)
Female sex	41 (70.9)	91 (53.2)	77 (55.0)	29 (42.6)	94 (52.5)	11 (45.8)
Diabetes
Total N	1510 (100)	3678 (100)	3263 (100)	1131 (100)	3807 (100)	353 (100)
Age, year, mean (SD)	55.7 (0.10)	56.0 (0.04)	56.2 (0.04)	55.5 (0.09)	56.1 (0.04)	54.9 (0.23)
Female sex	987 (65.4)	2303 (62.6)	2073 (63.5)	676 (59.8)	2388 (62.7)	205 (58.1)

Abbreviations. DM, disease management; CHD, coronary heart disease; HTN, hypertension; COPD, chronic obstructive pulmonary disease; CHF, congestive heart failure; CKD, chronic kidney disease.

^a The total number of patients varied year to year because members could become eligible on a rolling basis, and members could be counted twice in the year they switched from the “OUT” group to the “IN” group.

^b All values presented as no. (%) unless otherwise indicated.

^c By the last month of year 3, all members were actively enrolled in one of the DM programs. Therefore, the characteristics of the OUT cohort in year 3 are aggregated over 11 months (the last month of year 3 was excluded).

Descriptive evidence: individual programs

Costs

Table 3 describes unadjusted trends in average PMPM costs aggregated annually for IN and OUT patients for each of the 4 DM programs. In all 4 programs, average costs PMPM trended upward over time for the OUT cohort, while they remained relatively stable for the IN cohort. The CHD/HTN DM program was the only program to show lower average PMPM costs in the IN cohort relative to the OUT cohort in year 1. This trend was not observed until year 2 in the other 3 programs, which may be a reflection of the general acuity of CHD/HTN and its responsiveness to intervention.

Table 3. Average unadjusted PMPM costs per DM Program per year^a.

DM Program	Year 1		Year 2		Year 3
	IN	OUT	IN	OUT	IN	OUT
Diabetes	$1,141	$979	$1,151	$1,463	$1,265	$1,770
CHD/HTN	$1,154	$1,183	$931	$1,266	$977	$2,250
Asthma/COPD	$1,413	$1,316	$1,367	$2,542	$1,531	$3,892
CHF/CKD	$3,477	$2,483	$3,426	$4,169	$3,883	$5,063

Abbreviations. PMPM, per member per month; DM, disease management; CHD, coronary heart disease; HTN, hypertension; COPD, chronic obstructive pulmonary disease; CHF, congestive heart failure; CKD, chronic kidney disease.

^a Average unadjusted PMPM costs were aggregated annually to account for the decreasing sample size associated with the stepped-wedge study design. PMPM data collected for each patient within a 1-month wash-out period before and after initial DM engagement were excluded from analysis to reduce regression to mean.

The most pronounced separation in average PMPM costs for the OUT cohort compared to the IN cohort was observed in the asthma/COPD program. The highest average PMPM costs for both the IN and OUT cohorts were seen in the CHF/CKD program, which may reflect the seriousness of these 2 conditions. The average PMPM costs in the OUT cohorts for both the asthma/COPD and the CHF/CKD programs increased the most from year 1 to year 3.

Clinical impacts

DM programs improved clinical outcomes as well (Table 4). HbA1c readings showed an increasing trend for the diabetes OUT group and remained relatively stable for the IN group. Conversely, systolic BP values were fairly consistent in both IN and OUT cohorts over time; however, the average PMPM values were higher overall for patients OUT of the HTN/CHD program each year. Average PMPM diastolic BP values were only slightly elevated for years 1–3 in the OUT cohort and for years 1–2 in the IN cohort. Average diastolic values in year 3 were within the normal range for patients IN the HTN/CHD program.

Table 4. Average unadjusted clinical outcome measures per program per year^a.

Clinical outcome	Year 1		Year 2		Year 3
	IN	OUT	IN	OUT	IN	OUT
HbA1c, %	7.53	7.53	7.51	7.80	7.65	8.27
Systolic BP, mmHg	132.74	133.95	129.54	139.95	130.50	136.38
Diastolic BP, mmHg	81.85	81.79	82.74	84.30	75.50	82.13

Abbreviations. HbA1c, hemoglobin A1c; BP, blood pressure.

^a Average unadjusted clinical outcome measures were aggregated annually to account for the decreasing sample size associated with the stepped-wedge study design. Data collected for each patient within a 1-month wash-out period before and after initial DM engagement were excluded from analysis to reduce regression to mean.

Effects on medical utilization

Some DM programs also reduced acute and inpatient care. More specifically, emergency department (ED) visits, hospital admissions, and admission days were reduced for diabetes and HTN/CHD program members. In contrast, prescription medication use increased for diabetes, CHD/HTN, and CHF/CKD members.

Multivariable model evidence: combined programs

Multivariable regression model results were combined for all programs (Figure 1). As indicated in Figure 1, blended program savings for members in the program for 1, 2, and 3 years were −$31.61 (95% confidence interval [CI]: −$60.73, −$2.49), −$50.45 (95% CI; −$46.48, −$54.41), and −$53.72 (95% CI: −$43.54, −$63.90), respectively. When these costs are aggregated, a total estimated annual gross savings of $4.8 million was observed in diabetes, CHD/HTN, CHF/CKD, and asthma/COPD programs. The total 3-year cost gross savings was $14,460,174 (Table 5).

Figure 1. Regression model results: financial impacts of DM Programs. Abbreviations. DM, disease management; PMPM, per member per month; DiD, difference-in-difference. These are blended estimates, so model results represent medical PMPM trends and DiDs for all DM programs combined. Blended program savings for members in the program for 1, 2, and 3 years were −$31.61 (95% confidence interval [CI]: −$60.73, −$2.49), −$50.45 (95% CI: −$46.48, −$54.41), and −$53.72 (95% CI: −$43.54, −$63.90), respectively. When these costs are aggregated together, we estimate an annual savings of $4,820,058 based on the blended DiD of medical PMPM from the analysis.

Table 5. Medical savings by year.

Outcomes	1–12 months 3/1/2015–2/29/2016	13–24 months 3/1/2016–2/28/2017	25–36 months 3/1/2017–2/28/2018
“IN” versus “OUT” monthly trend difference	$32	$50	$54
Mean PMPM saving	$95	$177	$188
Mean “IN” months within that year	4.7	7.1	7.2
Mean PMPY saving	$446	$1,250	$1,358
Total members in DM within that year	2,178	4,719	5,591
Estimated total savings	$970,737	$5,899,630	$7,589,807
Total savings in 3 years	$14,460,174

Abbreviations. PMPM, per member per month; DM, disease management; PMPY, per member per year.

Discussion

This study investigated the impact of 4 DM programs implemented by BCBSLA for diabetes, CHD/HTN, asthma/COPD, and CHF/CKD. The study was conducted over 3 years on actively engaged members with medium- or high-acuity illnesses. The average gross savings for all programs, annualized by months of program participation, were estimated to be about $1,018 per participant per year, a finding that is consistent with the literature. For example, a systematic review by de Bruin et al. evaluated the effect of DM programs on healthcare costs for patients with diabetes (14 studies), depression (4 studies), HF (8 studies), or COPD (5 studies). Of the 31 total studies evaluated, 21 studies (67.7%) showed incremental costs that ranged between −$16,996 and $3,305 per patient per year for all disease groups, and 13 studies (41.9%) reported cost savings for DM program participants²².

The review by de Bruin and colleagues also postulated that DM programs for patients with COPD and HF might show cost savings earlier than DM programs for patients with diabetes because exacerbations associated with COPD and HF disease severity are more likely to result in expensive ED visits and hospitalizations²². Therefore, DM programs that are successful in reducing healthcare utilization will likely affect expenditures for patients with COPD or HF sooner. Similarly, the present study saw reasonably consistent PMPM costs over the study period for participants IN all of the BCBSLA DM programs, but the OUT cohort for the asthma/COPD and CHF/CKD programs had the largest increases in average PMPM costs from year 1 through year 3. This study also reported a reduction in ED visits and hospital admissions for participants in the diabetes and CHD/HTN programs; however, the DM program for CHD/HTN lowered costs in year 1, a trend that was not observed in the diabetes program (or any other program) until year 2.

While favorable results were not observed in year 1, the diabetes DM program was shown to be cost-effective in years 2 and 3. Many studies that evaluate DM programs, however, do not follow participants for more than 1 year. A 2012 study by Kogut et al. utilized data from Blue Cross Blue Shield of Rhode Island (BCBSRI) to evaluate the effects of a similar DM program for patients with diabetes²⁶. Like the BCBSLA DM program, participants in the BCBSRI DM program were offered a reduced copay for diabetic-related prescriptions and received individual support from a nurse and dietician. Twelve-month outcomes showed that patients with diabetes who participated in the BCBSRI DM program had higher per patient per year expenditures for prescription drugs than those who did not participate; however, the researchers did not observe a statistically significant difference between DM participants and non-participants for total prescription costs or all-cause medical care. The BCBSRI study focused on short-term results, whereas here, the focus is on higher-risk patients over a more extended period.

Other studies on DM programs for diabetes in the United States had longer follow-up times and reported comparable outcomes to this study. One analysis found that participation in a diabetes information system program (non-incentive-based) yielded average cost savings starting at about $504 per person after year 1 and reaching $3,563 in year 4 (program associated with a net reduction of $2,203 claims paid per patient per year after 32 months)²⁷. A 2018 paper looked at a multidisciplinary Risk Assessment and Management Program (RAMP) for diabetes and found the program to be cost-effective compared to usual primary care after 5 years (net savings of about $7,294 per participant)²⁸. Additionally, the prospective cohort study reported significantly lower complication rates for RAMP participants compared to non-participants (15.3% versus 28.7%, p < .001). Another study evaluated the effects of a chronic care model employed via telemedicine to improve outcomes for patients with diabetes and found that, at an average follow-up of 21 months, the intervention group also had significantly lower costs overall; however, the intervention neither improved clinical outcomes nor diabetes-related expenditures²⁹.

Conversely, a paper by Huang et al. showed that participation in a diabetes DM program through community health centers significantly improved several diabetes-related clinical outcomes, but costs ranged from $712 per patient in year 1 through $378 per patient in year 4³⁰. They ultimately determined, however, that such a program could be cost-effective if clinical improvements continued over time, which is seen in the literature. Better control of clinical factors associated with diabetes, like HbA1c and BP, has been shown to result in significant cost-savings, so DM programs that effectively reduce these measures will likely result in long-term savings^31,32. For example, a systematic review by Lian et al. assessed 8 cost-effectiveness studies for patients utilizing self-management programs for their Type 2 diabetes³³. Four of the studies that were evaluated showed cost savings ranging between $491 and $7,723 per unit reduction of HbA1c or body mass index. The present analysis showed improvements in HbA1c and systolic BP at 3 years for participants compared to non-participants. These promising results may be linked to the observed increases in prescription medication fills for those in the diabetes, CHD/HTN, and CHF/CKD DM programs and contributed to cost savings, especially given that HbA1C is not only a significant biomarker for diabetes, but also a reliable marker for coronary heart disease, stroke, dyslipidemia, and possibly HTN for patients with and without diabetes^34–37.

A 2-year retrospective analysis for an opt-in diabetes DM program also showed significant improvements in clinical measures like HbA1c and healthcare utilization³⁸. Furthermore, the study determined that annual savings for patients who were continuously enrolled in the diabetes DM program collectively amounted to about $4.0 million. Comparably, the 4 DM programs from BCBSLA combined saved about $4.8 million annually – a decisive finding within the context of the economic burden attributed to chronic disease in Louisiana. The number of people with multiple chronic illnesses is growing. Annual healthcare expenditures in the United States grew 3.9% to reach $3.5 trillion in 2017, and 90% of that spending was attributed to people with chronic and mental health conditions, with patients who had 5 or more chronic illnesses accounting for 41% of total spending but only making up 12% of the population^2,39,40. If trends continue, the projected medical and productivity costs of chronic disease in Louisiana per person per year are estimated to be $8,600 by 2030¹³. Cost-saving DM programs like the one implemented by BCBSLA, together with improved healthcare utilization and clinical outcomes, can help lessen the substantial clinical and economic burden of chronic illness in Louisiana and the United States.

Limitations

This study has some important limitations. The study design was non-experimental, so IN patients and OUT patients had some differences at baseline (sociodemographic information, sex, race, geographic location, employment status, etc.). Multivariable modeling was employed to address this issue; and, although these models adjusted for several clinical, demographic, and economic characteristics, it is possible that some important confounders were missing due to lack of data. Moreover, we acknowledge that the use of the stepped wedge design as an analytical methodology often relies on randomization and is not typically utilized outside of randomized controlled trials. Although our retrospective study was not a controlled trial, we felt the stepped wedge design was an appropriate methodology because it allowed us to maximize participant inclusion, as members become engaged in the DM program on a rolling basis. We also believe it helped control some selection bias because the entire study population was actively enrolled in and continuously engaged with a DM program by the end of year 3.

Another limitation was that cost-savings from the payer perspective were evaluated for all 4 DM programs combined due to sample size considerations and because members with multiple conditions were only enrolled in 1 DM program, so any disease-specific cost-savings could not be evaluated. However, all of these chronic conditions, including seemingly less acute conditions like asthma, lead to increased hospitalizations and healthcare utilization, especially in the absence of education or a management plan^41–45. Therefore, we believe our combined estimate for total medical costs adds value. Future studies with larger sample sizes may be able to conduct disease- or DM program-specific cost evaluations.

Additionally, passively enrolled members (those who were eligible for a DM program but did not actively participate) were omitted from the analysis. Because passively enrolled members were not evaluated, the study is potentially prone to selection bias, and it is difficult to assess the effectiveness of the DM programs’ ability to motivate members to actively participate. For this study, however, we were not looking at intent-to-treat effects. Instead, we were only interested in comparing those actively enrolled in a DM program to those not enrolled in a DM program because of the systematic differences between the groups (e.g. passive participants may be inherently healthier, or non-participants may be inherently less motivated).

Furthermore, some important clinical factors (e.g. smoking status as a marker for asthma/COPD disease progression) were not available. However, HbA1c was evaluated, which is not only a biomarker for diabetes but has also been shown to be an independent biomarker for cardiovascular disease, dyslipidemia, and possibly HTN^34–37. We plan to look at a more robust array of clinical factors for longer time frames in future studies.

Conclusions

DM programs implemented by BCBSLA to control diabetes, CHD/HTN, asthma/COPD, and CHF/CKD were economically effective, lowering costs with no evidence of worsening clinical outcomes; and, in some instances, even showing evidence of improved outcomes and healthcare utilization.

Transparency

Declaration of funding

This study was funded by Blue Cross Blue Shield of Louisiana.

Declaration of financial/other relationships

This research was performed internally by Blue Cross Blue Shield of Louisiana and authors neither received outside funding nor do they have any competing financial interests. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Author contributions

Miao Liu, Jason Ouyang, Yuan Zhang, and Xiaojing Yuan were involved in the conception and study design, data interpretation, and contributed to the interpretation of the results. Jason Ouyang, Yuan Zhang, and Somesh Nigam assisted with data interpretation and critical revisions. Janet Chaisson, Tasha Bergeron, Dawn Cantrell, and Vindell Washington own the DM programs and provided clinical and operational details about each interventional program. All authors provided critical feedback and helped shape the research and analysis.

Previous presentations

A version of this research was presented as a poster at the 2019 Blue Cross Blue Shield National Summit in Grapevine, TX, and the 2019 International Society for Pharmacoeconomics and Outcomes Research (ISPOR) conference in New Orleans, LO.

Acknowledgements

No assistance in the preparation of this article is to be declared.