Non–Face-to-Face Care Management and Service Utilization in Patients With Diabetes

ABSTRACT

Objectives: In 2015, CMS implemented reimbursement for non–face-to-face chronic care management (NFFCCM) for beneficiaries with multiple chronic conditions, including diabetes. This analysis estimated the association between NFFCCM and utilization of inpatient, outpatient, and emergency services.

Study Design: We implemented a doubly robust estimator using propensity score matching in a regression context to compare eligible patients who used NFFCCM (n = 282) with eligible patients who did not use NFFCCM (n = 26,759).

Methods: We tested 4 definitions of treatment: having any NFFCCM encounters and having 1 NFFCCM encounter per month, per 2 months, and per 3 months. Two-tailed statistical inference testing was conducted at the 5% level. We examined the utilization differences among patients with diabetes 65 years and older using merged electronic health records for 4 health systems in Louisiana from the Research Action for Health Network database in 2013 through 2018.

Results: We found NFFCCM was associated with increased utilization of care in the outpatient setting by 0.056 visits per month (95% CI, 0.027-0.086) and with lower utilization in the inpatient setting (–0.024 visits per month; 95% CI, –0.038 to –0.010) and in the emergency department setting (–0.017 visits per month; 95% CI, –0.031 to –0.003). Alternative specifications of treatment showed similar associations.

Conclusions: CMS implementation of reimbursement codes for NFFCCM, and subsequent utilization of that reimbursement by health systems, was associated with a shift in patient utilization from high-cost settings (inpatient and emergency department) to a lower-cost setting (outpatient office).

Am J Manag Care. 2023;29(8):417-422. https://doi.org/10.37765/ajmc.2023.89407

Takeaway Points

Non–face-to-face chronic care management may be useful for clinicians looking for ways to avert expensive medical encounters in their Medicare-eligible population with 2 or more chronic conditions. Since 2015, CMS has provided reimbursement for these telehealth encounters. In Louisiana among populations with diabetes, accessing these care management services led to the following:

• An increase in outpatient visits
• A decrease in inpatient visits
• A decrease in emergency department encounters

With high mortality and accompanying high treatment costs, diabetes presents a major challenge for the US health system.¹ Diabetes frequently coexists with other chronic conditions, presenting further difficulties in coordinating care delivery and disease management.² Chronic care management, the process of coordinating care between patients and providers for patients with chronic conditions, is critical to those with diabetes. However, quality care becomes less feasible in usual care models as diabetes advances and therapies become increasingly complex.³ A patient-centered approach that bridges gaps along the care pathway by connecting patients to providers and promoting patient and caregiver engagement is necessary.⁴ Care coordination is such an approach and has shown promise in chronic care management.^5-7 However, prior randomized controlled trials have demonstrated that efforts that include in-person care coordination efforts have been largely ineffective at reducing treatment costs, with the cost of the care coordination program largely offsetting decreases in other medical expenditures generated by the coordination program.⁸ Non–face-to-face chronic care management (NFFCCM) presents an attractive alternative to in-person care coordination with potentially lower coordination costs. Although many correlational studies have shown associations between receiving NFFCCM and reduced utilization and medical expenditures, the evidence base has lacked experimental and quasi-experimental designs to provide stronger evidence.⁹ This study intends to address this research gap by providing rigorous, observational evidence on the effectiveness of NFFCCM at shifting care from higher-cost settings (eg, inpatient care, emergency departments [EDs]) to lower-cost primary care or outpatient settings for patients with diabetes who have multiple chronic conditions.

In January 2015, CMS introduced a billing code that allows qualified practitioners (physicians, physician assistants, nurse practitioners, clinical nurse specialists, and certified nurse midwives) working in primary care practices to be reimbursed for NFFCCM. Services provided during NFFCCM include recording of patient health information (demographics, problems, medications, and medication allergies), maintaining a care plan that includes treatment goals and planned interventions, managing transitions of care, and sharing patient health information with providers outside the billing practice. This code allows practitioners to bill Medicare for NFFCCM once per month provided that the Medicare enrollee has 2 or more chronic conditions, the enrollee was provided at least 20 minutes of clinical staff time, the enrollee has a comprehensive care plan, and the practitioner providing the NFFCCM is either a physician or other qualified health care professional or supervised by such a professional. In general, clinics schedule a monthly call with the patient as a preventive check-in, with additional calls if needed (eg, for medication refills). Patients were also provided with a contact at the practice, typically a nurse, whom they could call as needed. The implementation of this code incentivized clinicians and health systems to implement NFFCCM and allowed us to track which patients received the service. We used a rigorous, observational design to analyze whether NFFCCM receipt was associated with changes in utilization of inpatient, outpatient, and emergency services.

METHODS

We implemented a doubly robust estimator¹⁰ using propensity score matching¹¹ in a regression context to compare eligible patients who used NFFCCM with eligible patients who did not use NFFCCM. Singly robust estimators (eg, propensity matching and regression) require that the statistical model be correctly specified. Doubly robust estimators provide an improvement because only 1 component of the estimator needs to be correctly specified to obtain an unbiased result.¹⁰ We obtained data on all patients with a diabetes diagnosis from the Research Action for Health Network (REACHnet) database between 2013 and 2018. REACHnet houses electronic health record data, standardized to the PCORnet Common Data Model,¹² from several health systems in Louisiana and Texas. For this study, we restricted the sample to patients 65 years and older because providers were eligible to receive reimbursement for NFFCCM services through Medicare’s new billing code.

Treated observations (n = 282) were assigned treatment at the date on which NFFCCM was first coded in their health record. To account for the variation in the timing of uptake of NFFCCM in the treatment group, untreated observations (n = 26,759) were randomly assigned initiation dates to match the distribution of initiation dates in the treated population to maximize the probability of matching successfully. Propensity scores were estimated using logistic regression based on baseline characteristics: sex, race (Black, White, and other), ethnicity (coded in 2 binary variables: Hispanic and whether Hispanic was missing), age in years at receipt of first NFFCCM, presence of chronic conditions other than diabetes (an indicator for the presence of a diagnosis of each of the following: Alzheimer disease, arthritis, asthma, atrial fibrillation, autism, cancer, chronic heart disease, chronic obstructive pulmonary disease, end-stage renal disease, depression, heart failure, hyperlipidemia, hypertension, osteoporosis, stroke, and schizophrenia (see the eAppendix Note [eAppendix available at ajmc.com] for International Classification of Diseases, Ninth Revision and Tenth Revision codes), whether Medicare was the primary payer for any claims at baseline, utilization per month at baseline (ie, hemoglobin A_1c tests, ED visits, inpatient visits, outpatient visits, observational stays, and other ambulatory visits), and number of months in the sample before treatment. Baseline measures started in 2013 and continued until the first month of treatment (or assigned treatment date for patients in the control), which means that each observation had at least 2 years of pretreatment data and, for late adopters, as many as 4 years of pretreatment data. To obtain a successful balance across treatment and control samples with the variable that indicated having Medicare as the primary payer at baseline, we also included interaction terms between that variable and the number of months in the sample before treatment, age in years at receipt of first NFFCCM, and sex. Treated observations were assigned unitary weights, and control observations were weighted by the odds (eg, pscore / [1 – pscore]).

To implement the doubly robust estimator, we used this sample in a weighted linear regression model. The regression model also controls for each of the variables used to obtain the matched sample, including the interaction terms. In the main model, the coefficient of interest returns the marginal association between ever having received NFFCCM and outpatient, inpatient, and ED visits per month. We also tested 3 other definitions of treatment in an effort to capture whether the frequency of NFFCCM was important. We defined treatment alternatively as having at least 1 NFFCCM encounter per month, having at least 1 NFFCCM encounter per 2 months, and having at least 1 NFFCCM encounter per 3 months. We estimated heterogenous treatment associations by age, race, and gender by adding an interaction term between an indicator for receipt of any NFFCCM visits and the indicated demographic characteristic to the regression models. Two-tailed statistical inference testing was conducted at the 5% level. All analyses were performed using Stata 15.1 (StataCorp LLC). The study and analysis plan were approved by the Tulane University, Pennington Biomedical Research Center, and Ochsner Health System institutional review boards.

RESULTS

Characteristics of the full sample are shown in the first 3 columns of Table 1. The last 3 columns show those characteristics after the matching and reweighting procedure. We successfully matched all baseline characteristics within 10% of a standardized difference for 282 treated observations and 26,759 control observations. Matches between the treatment and control groups on nonbinary variables are illustrated with histograms in eAppendix Figure 1 and eAppendix Figure 2.

Patients who received at least 1 NFFCCM encounter during the study period experienced fewer ED visits and inpatient stays after the NFFCCM encounter and increased outpatient visits after the NFFCCM encounter (Table 2). ED visits decreased by 0.017 (P < .05) per patient per month, inpatient stays decreased by 0.024 (P < .01) per patient per month, and outpatient visits increased by 0.056 (P < .05) per patient per month.

In Table 3, we explore the association between alternate measures of treatment and our 3 utilization measures. In the first row, receiving at least 1 NFFCCM encounter per 3 months was associated with a decline in ED visits of 0.015 (P < .05), a decline in inpatient visits of 0.022 (P < .01), and an increase in outpatient visits of 0.093 (P < .01) per patient per month. In the second row, receiving at least 1 NFFCCM encounter per 2 months was associated with a decline in inpatient visits of 0.02 (P < .05) and an increase in outpatient visits of 0.099 (P < .01) per patient per month. In the third row, receiving at least 1 NFFCCM encounter every month was associated with an increase in outpatient visits of 0.163 (P < .01) per patient per month. For each of our definitions of treatment, Wald tests could not rule out similar association sizes on ED visits or inpatient stays. There were significant differences in patient visits across the treatment definitions with more frequent NFFCCM being associated with more outpatient visits. Statistically significant differences in outpatient visit frequency were detected between the middle 2 frequencies (having 1 encounter per 3 months and 1 encounter per 2 months) and either ever having had an NFFCCM encounter or having an encounter every month.

Estimates of potential heterogenous treatment associations are presented in Table 4. We found no evidence of differential associations of NFFCCM by age (≤ 70 vs > 70 years) on any encounter type. We found no differential associations for ED visits or inpatient stays between men and women. We did find that NFFCCM increased the number of outpatient visits among women by 0.087 (P < .001) visits per month more than among men. We found no significant difference in the association between NFFCCM in Black patients (vs patients of other races) and ED visits. We did find that the decline in inpatient stays was attenuated among Black patients. The coefficient for inpatient stays on the interaction term was positive and statistically significant (0.033; P < .05), but because it was smaller in magnitude than the uninteracted treatment association (–0.048; P < .001) the association among Black patients was still a decrease. We did find that the increase in outpatient visits associated with NFFCCM was largely driven by the Black population (0.133; P < .001).

DISCUSSION

Patients who received NFFCCM experienced fewer ED visits, fewer inpatient stays, and increased outpatient visits after the NFFCCM encounter. Our core estimates indicate a decrease of 20.4 (0.017 × 12 × 100) ED visits per 100 patients annually and a decrease of 28.8 (0.024 × 12 × 100) inpatient stays per 100 patients annually. NFFCCM appears to help shift patient care from the ED and inpatient settings to outpatient visits, which is a major goal when attempting to provide chronic disease care in settings with lower, rather than higher, cost. We also examined whether patient service utilization rates were associated with different frequency levels of NFFCCM. We found similar impacts when we modeled treatment receipt at different intervals (at least once every 3 months, at least once every 2 months, or once a month). Generally, we could not statistically distinguish between the size of the treatment associations in these alternative models and the main model. This may imply that the first NFFCCM encounter is the most important and that regular NFFCCM encounters provided little additional benefit. However, because NFFCCM is a relatively recently initiated program, the sample we have for analysis may not be sufficient to show statistically different impacts across frequency levels of NFFCCM. Qualitative interviews with participants of the NFFCCM program have found that information about free transportation to regular appointments was especially helpful,^13,14 and it may only take a single NFFCCM encounter to impart this information.

We found no evidence of differential treatment associations by age. We did find that the increase in outpatient visits because of NFFCCM are largely driven by female or Black patients. These results may be due to marginalized populations, who are less attached to the medical system, being provided with information that connected them with the opportunity for outpatient visits. Further work is necessarily to identify heterogenous treatment associations by interactions across these demographics. As the program continues, more patients will utilize NFFCCM, which will increase the sample enough to allow for further examination.

Regardless, our finding of increased outpatient visits due to NFFCCM is present across several model specifications and subpopulations and is accompanied by slightly less robust decreases in patient stays and ED visits. Care for patients with diabetic conditions may be provided in each of these 3 modalities and may serve as substitutes for one another. Prior work has found that inpatient visits and outpatient visits may be substitutes for patient care across a variety of conditions.¹⁵ A patient with diabetes may be able to successfully manage their condition with outpatient visits and avoid the need for an ED visit or inpatient stay.¹⁶ NFFCCM may facilitate this substitution by providing patients with a care treatment plan, coordination of multiple providers, easier medication adherence, and information about transportation to regular clinic visits.

Strengths and Limitations

This is the first study to examine the impact of NFFCCM on utilization of inpatient, outpatient, and emergency services among patients with diabetes in Louisiana. We applied a doubly robust estimator to remove any of the selection bias correlated with observables. There are, however, important limitations to this study. First, participants in NFFCCM were not randomly assigned. Doctors may have recruited patients most likely to benefit from the program, and patients most likely to benefit may have agreed to participate. To the extent that the likeliness to benefit is not related to observable characteristics (eg, presence of other chronic conditions, utilization at baseline, age), then our results may be biased toward finding larger impacts. Patients may have experienced other chronic disease management programs concurrently with receiving NFFCCM. If enrollment in these other programs was correlated with enrollment in NFFCCM, we may have captured the combined effect of NFFCCM and other programs. Second, these results are from a specific population in Louisiana limited to a few health systems. Extrapolating from the experience here to predicting how the program will perform in other health systems may require more detailed knowledge of how NFFCCM encounters are conducted. Third, the end points examined here are utilization measures rather than health outcomes. Although fewer ED visits and inpatient stays may be a positive indicator for patient health, we did not examine patient health outcomes in this study. Fourth, 2 years after the introduction of reimbursement code 99490 described in the introduction, CMS introduced 3 additional codes (G0506, 99487, and 99490) to reimburse additional staff time for more complex cases. In our data, these additional codes appeared rarely (32 instances in 2018). Because the take-up of these new codes was rare, we have not distinguished between reimbursement for NFFCCM under the initial 99490 code or the new codes. Future work could examine differential impacts of the new codes if uptake continues.

CONCLUSIONS

This study finds small but statistically significant associations between NFFCCM usage and declines in ED and inpatient visits and increases in outpatient visits among older patients with diabetes. These may reflect patients with diabetes being better able to manage their disease with support provided through remote care coordination. Further studies examining the impact of these NFFCCM encounters on diabetic outcomes, such as hemoglobin A_1c levels, and total economic cost are warranted.

Acknowledgments

The content of this manuscript is solely the responsibility of the author(s) and does not necessarily represent the views of the Patient-Centered Outcomes Research Institute. The study team would like to acknowledge the contributions of our partners: Ochsner Health System; Tulane Medical Center; University Medical Center New Orleans; EXCELth, Inc; DePaul Community Health Centers; Access Health Louisiana; Research Action for Health Network (REACHnet, a PCORnet Clinical Research Network) and their multistakeholder Diabetes Advisory Groups; Pennington Biomedical Research Center; Blue Cross and Blue Shield of Louisiana; and our patient partners, Patricia Dominick and Cathy Glover. The authors also thank colleagues in the NEXT-D2 study network for invaluable feedback.

Author Affiliations: Department of Health Policy and Management (CS, LS) and Department of Global Community Health and Behavioral Sciences (ANB), Tulane University School of Public Health and Tropical Medicine, New Orleans, LA; Louisiana Public Health Institute (EN), New Orleans, LA; Ochsner Health Center for Outcomes and Health Services Research (EGP-H), New Orleans, LA; Section of Endocrinology, Department of Medicine, Tulane University (YY), New Orleans, LA.

Source of Funding: This research was supported by the Patient-Centered Outcomes Research Institute (PCORI) (NEN-1508-32257).

Author Disclosures: Dr Neuman reports employment with the Louisiana Public Health Institute and grants received from PCORI. Dr Price-Haywood reports appointment to the PCORI board of governors after this article was completed. Dr Shi was the principal investigator on the PCORI grant that funded this study, which paid a portion of all the authors’ salaries to their respective employers, and has also received unrelated funding from PCORI.

Authorship Information: Concept and design (CS, EN, YY, LS); acquisition of data (CS, EN, EGP-H, LS); analysis and interpretation of data (CS, EN, ANB, EGP-H, YY, LS); drafting of the manuscript (CS, EGP-H, LS); critical revision of the manuscript for important intellectual content (CS, EN, ANB, YY, LS); statistical analysis (CS); provision of patients or study materials (EN); obtaining funding (CS, EN, LS); administrative, technical, or logistic support (EN, ANB); and supervision (LS).

Address Correspondence to: Charles Stoecker, PhD, Department of Health Policy and Management, Tulane University School of Public Health and Tropical Medicine, 1440 Canal St #8346, New Orleans, LA 70118. Email: cfstoecker@tulane.edu.

REFERENCES

1. Menke A, Casagrande S, Geiss L, Cowie CC. Prevalence of and trends in diabetes among adults in the United States, 1988-2012. JAMA. 2015;314(10):1021-1029. doi:10.1001/jama.2015.10029

2. Piette JD, Kerr EA. The impact of comorbid chronic conditions on diabetes care. Diabetes Care. 2006;29(3):725-731. doi:10.2337/diacare.29.03.06.dc05-2078

3. Huynh P, Toulouse A, Hirsch IB. One-year time analysis in an academic diabetes clinic: quantifying our burden. Endocr Pract. 2018;24(5):489-491. doi:10.4158/EP-2017-0267

4. Yeager VA, Wharton MK, Monnette A, et al. Non-face-to-face chronic care management: a qualitative study assessing the implementation of a new CMS reimbursement strategy. Popul Health Manag. 2018;21(6):454-461. doi:10.1089/pop.2017.0196

5. Hess BJ, Lynn LA, Holmboe ES, Lipner RS. Toward better care coordination through improved communication with referring physicians. Acad Med. 2009;84(suppl 10):S109-S112. doi:10.1097/ACM.0b013e3181b37ac7

6. O’Malley AS, Tynan A, Cohen GR, Kemper N, Davis MM. Coordination of care by primary care practices: strategies, lessons and implications. Res Brief. 2009;(12):1-16.

7. Pham HH, O’Malley AS, Bach PB, Saiontz-Martinez C, Schrag D. Primary care physicians’ links to other physicians through Medicare patients: the scope of care coordination. Ann Intern Med. 2009;150(4):236-242. doi:10.7326/0003-4819-150-4-200902170-00004

8. Peikes D, Chen A, Schore J, Brown R. Effects of care coordination on hospitalization, quality of care, and health care expenditures among Medicare beneficiaries: 15 randomized trials. JAMA. 2009;301(6):603-618. doi:10.1001/jama.2009.126

9. Verhoeven F, Tanja-Dijkstra K, Nijland N, Eysenbach G, van Gemert-Pijnen L. Asynchronous and synchronous teleconsultation for diabetes care: a systematic literature review. J Diabetes Sci Technol. 2010;4(3):666-684. doi:10.1177/193229681000400323

10. Funk MJ, Westreich D, Wiesen C, Stürmer T, Brookhart MA, Davidian M. Doubly robust estimation of causal effects. Am J Epidemiol. 2011;173(7):761-767. doi:10.1093/aje/kwq439

11. Garrido MM, Kelley AS, Paris J, et al. Methods for constructing and assessing propensity scores. Health Serv Res. 2014;49(5):1701-1720. doi:10.1111/1475-6773.12182

12. Fleurence RL, Curtis LH, Califf RM, Platt R, Selby JV, Brown JS. Launching PCORnet, a national patient-centered clinical research network. J Am Med Inform Assoc. 2014;21(4):578-582. doi:10.1136/amiajnl-2014-002747

13. Bazzano A, Monnette A, Wharton M, et al. Older patients’ preferences and views related to non-face-to-face diabetes chronic care management: a qualitative study from southeast Louisiana. Patient Prefer Adherence. 2019;13:901-911. doi:10.2147/PPA.S201072

14. Bazzano AN, Wharton MK, Monnette A, et al. Barriers and facilitators in implementing non-face-to-face chronic care management in an elderly population with diabetes: a qualitative study of physician and health system perspectives. J Clin Med. 2018;7(11):451. doi:10.3390/jcm7110451

15. Fortney JC, Steffick DE, Burgess JF Jr, Maciejewski ML, Petersen LA. Are primary care services a substitute or complement for specialty and inpatient services? Health Serv Res. 2005;40(5, part 1):1422-1442. doi:10.1111/j.1475-6773.2005.00424.x

16. Chung N, Rascati K, Lopez D, Jokerst J, Garza A. Impact of a clinical pharmacy program on changes in hemoglobin A1c, diabetes-related hospitalizations, and diabetes-related emergency department visits for patients with diabetes in an underserved population. J Manag Care Spec Pharm. 2014;20(9):914-919. doi:10.18553/jmcp.2014.20.9.914