John D. Piette, PhD

 To assess what patient, family supporter, and call characteristics predicted whether patients completed automated and coach-provided calls in a telehealth diabetes intervention.

 A total of 123 adults with type 2 diabetes and high glycated hemoglobin A

) or blood pressure, enrolled with a family supporter, received automated interactive voice response (IVR) and coach-provided visit preparation calls over 12 months.

 Data from baseline surveys and diabetes-related clinical information from patient medical records were entered into multilevel, multivariate regression models of associations between participant and call characteristics with call completion.

 A total of 76.3% of 2784 IVR calls and 75.8% of 367 visit preparation calls were completed. For IVR calls, patients with recent call-triggered provider alerts had higher odds of call completion (adjusted odds ratio [AOR], 3.5; 95% CI, 2.2-5.5); those with depressive symptoms (AOR, 0.4; 95% CI, 0.2-0.9), higher HbA

 (AOR, 0.8; 95% CI, 0.6-0.99), and more months in the study (AOR, 0.9; 95% CI, 0.87-0.94 per month) had lower odds. For visit preparation calls, higher patient activation scores predicted higher call completion (AOR, 1.4; 95% CI, 1.1-1.9); patient college education predicted less call completion (AOR, 0.3; 95% CI, 0.2-0.6). Supporter help taking medications predicted less completion of both call types. Patient age did not predict call completion.

 Patients of all ages completed telehealth calls at a high rate. Automated IVR calls were completed more often when urgent issues were identified to patients’ providers, but less often if patients had high HbA

 or depression. Visit preparation call content should be tailored to patient education level. Family help with medications may identify patients needing additional support to engage with telehealth.

More than 34 million adults in the United States have type 2 diabetes,

 and successful management requires frequent patient follow-up to assess efficacy of treatment regimens, support self-care behavior, and detect emerging problems. Telehealth programs are an increasingly popular approach to facilitate this care by enhancing remote monitoring, communication, and collaboration among patients, caregivers, and health care providers.

 Interactive voice response (IVR) calls, automated telephone calls that can provide tailored patient monitoring and self-care education, are one form of telehealth intervention that can facilitate self-care support between face-to-face visits and result in lower glycated hemoglobin A

 IVR calls have some advantages over other forms of telehealth that involve smartphone apps or specialized devices because they can be used with any type of phone and thus present lower barriers relating to literacy, comfort with technology, internet connection, and smartphone or computer availability.

 As a result, IVR has been used successfully in diverse populations including older adults, Spanish-speaking patients, and veterans.

 However, little is known about which patients are most likely to engage with IVR and how IVR influences patient-provider interactions.

In addition to routine between-visit monitoring, using telehealth in certain key situations, such as prior to scheduled medical visits, can have particularly high impact on diabetes management. For patients with diabetes, visit preparation phone calls and messaging have been shown to significantly improve engagement in treatment decisions and adherence to lipid and HbA

 To enhance use of phone-based telehealth for visit preparation, we need to gain a better understanding of patient and visit factors that affect patient engagement with these calls.

One key factor that may enhance patient engagement in telehealth is family or caregiver support. Multiple studies have demonstrated that patients who enroll with a family supporter are more likely to answer intervention calls and have greater improvement in monitored behaviors than patients who enroll alone.

 IVR calls with caregiver feedback can also improve patient medication adherence and reduce diabetes distress.

 The specific effects of family supporter characteristics and diabetes support roles on how patients engage with phone-based care have previously not been studied.

In this study, we examined which patient, family supporter, and call characteristics predicted engagement in calls conducted as part of a telehealth intervention with adults with type 2 diabetes. The intervention used 2 types of calls. Automated biweekly IVR calls monitored for significant variations in glucose and blood pressure self-monitoring, symptoms, and issues with medication adherence, and prompted patient, family, and primary care provider actions when appropriate. Visit preparation calls conducted live by health coaches facilitated patient preparation for and engagement in scheduled provider visits. Our specific research questions were: (1) Which patient, family supporter, and call characteristics predicted completion of automated IVR calls?; (2) Which patient, family supporter, and call characteristics predicted completion of coach-patient visit preparation calls?; and (3) How did health care providers and patients respond to receiving automated IVR calls to monitor diabetes and the urgent alerts they generated?

In the Caring Others Increasing EngageMent in Patient-Aligned Care Teams (CO-IMPACT) study, 123 adult patients—each with type 2 diabetes, specific risk factors for complications, and 1 family supporter—received an IVR and visit preparation call intervention over 12 months. Patients were veterans receiving primary care at 2 Veterans Health Administration sites, which provided comprehensive diabetes care with a primary care provider, clinical pharmacist, nutritionist, and diabetes self-management classes. This study was approved by the Veterans Health Administration Ann Arbor Institutional Review Board. Study enrollment began in November 2016 and final participant data collection occurred in June 2019.

Patients had to have a diagnosis of type 2 diabetes and have poor glycemic (HbA

 > 8%) or blood pressure (systolic blood pressure [SBP] > 160 mm Hg) control according to their most recent measurement in the previous 9 months. To identify a family supporter, patients were asked, “Do you have a family member or friend who gets involved with your health care in one of these ways…” followed by a list of specific support roles. Supporters could live with or apart from the patient but had to speak with the patient at least twice per month about their health issues or health care. Additional details on patient and health supporter eligibility, enrollment, and source study design can be found in the published study protocol.

Patient-supporter dyads received 3 components: an initial coaching session, biweekly IVR calls, and coach-delivered visit preparation phone calls (

). IVR calls allowed patients to report and receive information about diabetes- and hypertension-related concerns by responding to prerecorded prompts. Family supporters received summaries of issues identified in calls and information on how they could help with these concerns. When patients indicated they had hyperglycemia (glucose > 300 mg/dL), hypoglycemia (glucose < 70 mg/dL) or symptoms, hypertension (SBP > 170 mm Hg), or hypotension (SBP < 90 mm Hg) in IVR calls, urgent faxes were sent to the patient’s primary care team notifying them of the patient’s responses. Patients were notified at the time of the call that a message would be sent to their primary care team and had the option of being connected directly to the clinic via phone. Health coaches, who had a background in health education and received training in basic diabetes management, autonomy-supportive communication, and concepts underlying patient activation and action planning, made visit preparation calls prior to primary care visits and focused on identifying diabetes-related concerns and bringing self-monitoring results to appointments.

Our primary outcome measure was whether the patient answered a call during a time period eligible for call receipt.

 Patients were eligible to receive an IVR call once in every 2-week period. The IVR system attempted to call the participant 3 times during this period, and times during which the patient indicated they would be unavailable (eg, for vacation) were excluded.

 Visit preparation calls were attempted 1 week prior to a primary care provider, nurse, or clinical pharmacist visit. If a preparation call was completed but the patient did not attend the visit, or if a call was not attempted (eg, if the appointment was made last minute), that call or visit was excluded from analyses.

Sociodemographic and medical data were collected by patient survey at baseline. Blood pressure and HbA

 measurements were collected during this baseline assessment.

 Baseline patient measures included patient activation (Patient Activation Measure-13 [PAM-13]

), activation in health care appointments (Perceived Efficacy in Patient-Physician Interactions

), diabetes-related distress (Problem Areas In Diabetes scale

), and satisfaction with primary care team (1 item from Consumer Assessment of Healthcare Providers and Systems survey

). Variables related to patient–family supporter relationship and communication included level of autonomy support perceived by the patient, measured by the Important Other Climate Questionnaire,

 and items measuring family supporter involvement in specific aspects of diabetes care and helpfulness, included in 

). The Brief Health Literacy Screen assessed patient health literacy and was dichotomized consistent with prior studies among veterans.

 The Patient Health Questionnaire screened for depressive symptoms and was considered positive for scores of 3 or greater.

 At the end of the 12-month intervention, patients rated how much the components of the intervention helped them manage their diabetes on a 4-point scale (1, helped me a great deal; 4, seemed to make things worse) and provided open-ended feedback on the intervention calls.

 Family supporters completed baseline surveys assessing their own sociodemographic and medical characteristics and health literacy.

 For visit preparation calls, we included months in study at time of call and the type of provider for the clinic visit. For IVR calls, we included months in study at time of call, call completion, and urgent fax receipt in the prior 4 weeks.

Patient, supporter, and call characteristics were examined in 2 separate mixed effects regression models designed to predict IVR and visit preparation call completion. To adjust for the effect of varying call completion rates between and within dyads, each model was fit to predict call completion while incorporating random effects for dyad-level characteristics. To assess the effect of recent urgent fax on IVR call completion while accounting for confounding with recent call completion, a trilevel variable of prior 4-week call completion by fax received was created. Only patients with nonmissing data were included (all variables missing ≤ 2%). After full initial models, parsimonious models were created through manual backward selection, removing variables at each step with a 

 value > .10, then retaining 3 of the most theoretically important variables from prior literature on diabetes self-management: age, insulin use, and baseline HbA

. This modeling approach balanced considerations of both clinical and statistical significance when choosing among the many predictor variables available, allowing us to compare effects of variables with known clinical relevance to diabetes care as well as discover novel predictors of engagement. Results of the final models are presented in main tables, and full initial model results are presented in 

. Data were analyzed using SAS 9.4 (SAS Institute Inc).

For each urgent fax generated, the electronic health record was reviewed to determine if the primary care team documented receipt of the fax, the role of the person responding to the fax (registered nurse, primary care provider, pharmacist), qualitative information on actions taken in response, and whether a primary care visit was scheduled to occur within 2 weeks. Patient ratings of intervention satisfaction are summarized descriptively. For open-ended participant feedback, informally selected quotes are used as context in the discussion of main quantitative results.

A total of 123 patient-supporter pairs were enrolled in the intervention and completed the baseline assessment. Of these, 8 patients and 15 family supporters died or withdrew prior to completing the 12 months of the intervention period, but any data collected on eligible calls prior to termination of study participation were included in analyses.

Patient and Family Supporter Characteristics

Of the 114 patients in visit preparation call models and 116 participants in IVR call models, 111 patient and supporter pairs were included in both. Baseline characteristics were similar between the 2 model cohorts (

Among 2784 eligible automated biweekly IVR call periods, 2124 (76.3%) calls were completed, with a median of 25 (range, 6-28) per patient over the 12-month intervention. Higher baseline HbA

 (adjusted odds ratio [AOR], 0.80; 95% CI, 0.64-0.99 per 1% increase), patient depressive symptoms (AOR, 0.36; 95% CI, 0.15-0.87), and family supporters assisting patients with taking medications (AOR, 0.45; 95% CI, 0.24-0.84) were associated with lower odds of call completion (

). More months in the study at time of IVR call was also associated with lower odds of call completion (AOR, 0.91; 95% CI, 0.87-0.94 per additional month). Having a call completed in the prior 4 weeks that did not trigger an urgent fax predicted current call completion (AOR, 2.87; 95% CI, 1.99-4.15 vs no recent call completion). Having a call completed in the prior 4 weeks that did trigger an urgent fax was associated with even higher odds of call completion (AOR, 3.42; 95% CI, 2.15-5.42 vs no recent call completion). There was no significant association between other variables and call completion (see eAppendix Table 2 for full model results).

Predictors of Visit Preparation Call Completion

Among 367 eligible primary care visits, 278 (75.8%) visit preparation calls were completed. A median of 2.0 (range, 0-13) visit preparation calls were completed per patient over the 12-month intervention period. Higher patient activation (AOR, 1.42; 95% CI, 1.09-1.85 per 10-point increase in PAM-13) predicted higher odds of call completion (

). College education predicted lower odds of call completion (AOR, 0.32; 95% CI, 0.18-0.57). As with IVR calls, the only family supporter factor significantly associated with call completion was assistance with taking medications (AOR, 0.52; 95% CI, 0.30-0.90). There was no significant association between other predictors and call completion (see eAppendix Table 3 for full model results).

Urgent Fax Responses and Patient Satisfaction

Of 116 participants with IVR call data, 76 had at least 1 urgent fax generated, with a mean of 2.9 faxes per participant over the 12-month period. The most common reasons for urgent faxes were hyperglycemia, high blood pressure, and hypoglycemia symptoms. For 47% of alerts, the fax was acknowledged in the patient’s chart or a primary care visit occurred within 2 weeks. The majority of faxes were acknowledged by the primary care nurse (60%), and responses included bringing the patient into clinic early for follow-up and delivering self-management advice or medication adjustment over the phone. At the end of the intervention, the majority of patients (69% for visit preparation and 75% for IVR calls) reported that the calls were helpful to their diabetes management.

With rapidly growing reliance on telehealth in diabetes care, the ability to identify and address issues limiting patient engagement with telehealth will be increasingly important. In this phone-based intervention with adults with type 2 diabetes and a family supporter, engagement was high for patients of all ages. Nearly half of urgent alerts to clinicians generated by IVR responses were acknowledged in the medical chart, facilitating identification of time-sensitive issues between visits. Consistent with prior literature, IVR calls were perceived as helpful by the majority of patients and were answered more than 75% of the time.

We identified several notable predictors of engagement with automated IVR calls. Our finding that patients with higher baseline HbA

 and more depressive symptoms were less likely to complete IVR calls is consistent with prior literature.

 This may indicate that these patients need additional support to facilitate IVR engagement, such as treatment of depression prior to automated call interventions or incorporation of depression symptom or medication monitoring into IVR calls.

 It is worth noting, however, that even if a patient responded to only half of the possible IVR calls, this would still result in 13 additional contacts during a 1-year period compared with the 4 face-to-face visits most patients have annually. Additionally, there was no association between patient age and IVR call completion, supporting the idea that telephone-based interventions can overcome barriers that older patients may face when engaging with other forms of virtual care.

Although our findings align with prior evidence that engagement with telehealth interventions tends to decrease over time,

 our finding that patients with recent urgent concerns flagged by IVR responses were more likely to complete calls demonstrates that the population with active issues may be more engaged in automated monitoring. Alerts to primary care teams may have fostered patients’ sense that IVR calls were useful, as patients had clear feedback that their responses were being communicated to their care teams. This sentiment was expressed in several interview responses from patients at the end of the study; for example, one patient noted: “If I reported something high or low my doctor would call me. It helped me but I also think it helped my doctor care for me better. I only see the doctor every 4 months and so if I had a bad 2 weeks, my doctor would call and have me adjust my medicine over the phone so I don’t have to come in.”

We also identified important patient factors that predicted visit preparation call engagement. Patient activation has been associated with increased engagement with other telehealth interventions.

 Our finding that college education was associated with less visit preparation call completion was unexpected, as prior literature has shown that higher education predicts more engagement in telemonitoring for diabetes.

 One possible explanation is that patients with higher education levels routinely perform visit preparation on their own and thus found visit preparation calls less useful. This sentiment was reflected by multiple patients at the end of the study; for example, one patient noted: “I already knew what to do, so when she called me, I told her, I already have that ready and the logs and know what I am going to ask the doctor.”

Multiple studies have demonstrated the beneficial effect of family supporters on patient engagement with telehealth interventions.

 In our study, we examined more nuanced family supporter behaviors to delineate which contributed to patients’ call engagement, and we surprisingly found that for both call types, patients who had support for taking medications had lower rates of call completion. One possible explanation is that this support role identifies patients who are more dependent on supporter help in general; alternatively, patients with high levels of in-person support may not feel the need to use technology to connect to supporters. Further research should determine what needs these patients have that would facilitate engagement with telehealth calls; it is possible that “live” monitoring calls that include both patients and supporters would be more beneficial for these patients than automated calls to patients alone.

There are limitations of this study that should be considered when interpreting our findings. Our sample included predominantly male veterans with limited racial and ethnic diversity, and the study was conducted in 2 primary care sites in 1 geographic region, all of which may limit generalizability of findings to other populations and settings. The smaller number of visit preparation calls may have limited statistical power to identify factors associated with completion of this type of call in the initial model given the number of covariates included.

Despite these limitations, our study has several important implications. In the midst of the COVID-19 pandemic and rapid growth of telehealth, there has been concern about marginalized populations, specifically elderly patients, who often have less technological literacy, being left behind with unmet care needs while confined to their homes.

 Our findings suggest that telephone-based interventions may be an effective way to augment monitoring of patients with diabetes, including older patients and those unable to complete video calls or engage with online portals. Our results indicate that diabetes telehealth programs will need to be carefully adapted to the needs of patients with poor glycemic control, with depression, and who require assistance with medication regimens. Automated call interventions may be more effective if they include a mechanism for feedback between clinicians and patients to encourage action on concerning diabetes-related issues and to reinforce the benefits of engagement with automated calls for patients.

Our findings indicate that adults with type 2 diabetes had high engagement with both automated and coach-delivered diabetes management calls over the span of 12 months and high satisfaction with this mode of care. Groups at increased risk of poor diabetes outcomes, including older patients and those with low health literacy, engaged with intervention calls at rates similar to groups without these risk factors. Optimizing the reach, engagement, and effectiveness of diabetes telehealth, particularly for those most vulnerable to poor diabetes outcomes, will only grow in importance as we work to improve the efficiency and quality of diabetes care in an increasingly technological era of care delivery.

 VA Pittsburgh Center for Health Equity Research and Promotion (MFZ, DSO, AOY, AMR), Pittsburgh, PA; Department of Medicine, University of Pittsburgh Medical School (MFZ, DSO, AMR), Pittsburgh, PA; VA Ann Arbor Center for Clinical Management Research (JDP, SCS, MH), Ann Arbor, MI; University of Michigan School of Public Health (JDP, SCS, MH), Ann Arbor, MI; VA Pittsburgh Health System Research Office (MB-K), Pittsburgh, PA; Department of Biostatistics, University of Pittsburgh Graduate School of Public Health (AOY), Pittsburgh, PA; University of Colorado (LO), Denver, CO; VA Palo Alto Health Care System, Center for Innovation to Implementation (RT), Palo Alto, CA; Division of Public Mental Health and Population Sciences, Stanford University (RT), Stanford, CA; University of Michigan Medical School (MH), Ann Arbor, MI.

 Veterans Health Administration Health Services Research and Development IIR 14–074-1; University of Pittsburgh T32 Research Training in Diabetes and Endocrinology 5T32DK007052-45; Michigan Center for Diabetes Translational Research (National Institutes of Health grant 5P60-DK09292).

 The authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

 Concept and design (MFZ, JDP, SCS, RT, MH, AMR); acquisition of data (SCS, DSO, MBK, LO, AMR); analysis and interpretation of data (MFZ, DSO, MB-K, AOY, LO, RT, AMR); drafting of the manuscript (MFZ, SCS, MB-K, AMR); critical revision of the manuscript for important intellectual content (MFZ, JDP, SCS, DSO, MB-K, AOY, LO, RT, MH, AMR); statistical analysis (MFZ, DSO, MB-K, AOY, AMR); provision of patients or study materials (JDP, SCS, RT, MH, AMR); obtaining funding (JDP, MH, AMR); administrative, technical, or logistic support (SCS, AMR); and supervision (JDP, AOY, MH, AMR).

 Margaret F. Zupa, MD, Department of Medicine, University of Pittsburgh, Falk Medical Bldg, 3601 Fifth Ave, Ste 3B, Pittsburgh, PA 15213. Email: zupamf@upmc.edu.

National Diabetes Statistics Report, 2020: Estimates of Diabetes and Its Burden in the United States

2. Appuswamy AV, Desimone ME. Managing diabetes in hard to reach populations: a review of telehealth interventions. 

. 2020;20(7):28. doi:10.1007/s11892-020-01310-2

3. Posadzki P, Mastellos N, Ryan R, et al. Automated telephone communication systems for preventive healthcare and management of long-term conditions. 

. 2016;12(12):CD009921. doi:10.1002/14651858.CD009921.pub2

4. Roberts ET, Mehrotra A. Assessment of disparities in digital access among Medicare beneficiaries and implications for telemedicine. 

. 2020;180(10):1386-1389. doi:10.1001/jamainternmed.2020.2666

5. Anderson M, Perrin A, Jiang J, Kumar M. 10% of Americans don’t use the internet. who are they? Pew Research Center. April 22, 2019. Accessed August 18, 2020. https://web.archive.org/web/20200818084042/https://www.pewresearch.org/fact-tank/2019/04/22/some-americans-dont-use-the-internet-who-are-they/

6. Magid DJ, Ho PM, Olson KL, et al. A multimodal blood pressure control intervention in 3 healthcare systems. 

7. Piette JD, Kerr E, Richardson C, Heisler M. Veterans Affairs research on health information technologies for diabetes self-management support. 

. 2008;2(1):15-23. doi:10.1177/193229680800200104

8. Piette JD, McPhee SJ, Weinberger M, Mah CA, Kraemer FB. Use of automated telephone disease management calls in an ethnically diverse sample of low-income patients with diabetes

. 1999;22(8):1302-1309. doi:10.2337/diacare.22.8.1302

9. Rivo J, Page TF, Arrieta A, et al. The impact of comprehensive pre-visit preparation on patient engagement and quality of care in a population of underserved patients with diabetes: evidence from the care management medical home center model

. 2016;19(3):171-177. doi:10.1089/pop.2015.0063

10. Vo MT, Uratsu CS, Estacio KR, et al. Prompting patients with poorly controlled diabetes to identify visit priorities before primary care visits: a pragmatic cluster randomized trial. 

. 2019;34(6):831-838. doi:10.1007/s11606-018-4756-4

11. Aikens JE, Trivedi R, Aron DC, Piette J. Integrating support persons into diabetes telemonitoring to improve self-management and medication adherence. 

. 2015;30(3):319-326. doi:10.1007/s11606-014-3101-9

12. Piette JD, Rosland AM, Marinec NS, Striplin D, Bernstein SJ, Silveira MJ. Engagement with automated patient monitoring and self-management support calls: experience with a thousand chronically ill patients. 

. 2013;51(3):216-223. doi:10.1097/MLR.0b013e318277ebf8

13. Aikens JE, Rosland AM, Piette JD. Improvements in illness self-management and psychological distress associated with telemonitoring support for adults with diabetes. 

. 2015;9(2):127-134. doi:10.1016/j.pcd.2014.06.003

14. Rosland AM, Piette JD, Trivedi R, et al. Engaging family supporters of adult patients with diabetes to improve clinical and patient-centered outcomes: study protocol for a randomized controlled trial. 

. 2018;19(1):394. doi:10.1186/s13063-018-2785-2

15. Hibbard JH, Stockard J, Mahoney ER, Tusler M. Development of the Patient Activation Measure (PAM): conceptualizing and measuring activation in patients and consumers. 

. 2004;39(4, pt 1):1005-1026. doi:10.1111/j.1475-6773.2004.00269.x

16. Maly RC, Frank JC, Marshall GN, DiMatteo MR, Reuben DB. Perceived efficacy in patient-physician interactions (PEPPI): validation of an instrument in older persons. 

. 1998;46(7):889-894. doi:10.1111/j.1532-5415.1998.tb02725.x

17. ten Klooster PM, Oostveen JCM, Zandbelt LC, et al. Further validation of the 5-item Perceived Efficacy in Patient–Physician Interactions (PEPPI-5) scale in patients with osteoarthritis. 

. 2012;87(1):125-130. doi:10.1016/j.pec.2011.07.017

18. McGuire BR, Morrison TG, Hermanns N, et al. Short-form measures of diabetes-related emotional distress: the Problem Areas in Diabetes Scale (PAID)-5 and PAID-1. 

. 2010;53(1):66-69. doi:10.1007/s00125-009-1559-5

19. CAHPS Clinician and Group Survey 3.0 measures. Agency for Healthcare Research and Quality. Updated February 2019. Accessed July 20, 2020. https://www.ahrq.gov/cahps/surveys-guidance/cg/about/survey-measures.html

20. Williams GC, Lynch MF, McGregor HA, Ryan RM, Sharp D, Deci EL. Validation of the “Important Other” Climate Questionnaire: assessing autonomy support for health-related change. 

. 2006;24(1):179-194. doi:10.1037/1091-7527.24.2.179

21. Chew LD, Griffin JM, Partin MR, et al. Validation of screening questions for limited health literacy in a large VA outpatient population. 

. 2008;23(5):561-566. doi:10.1007/s11606-008-0520-5

22. Löwe B, Kroenke K, Gräfe K. Detecting and monitoring depression with a two-item questionnaire (PHQ-2). 

. 2005;58(1):163-171. doi:10.1016/j.jpsychores.2004.09.006

23. Piette JD. Satisfaction with automated telephone disease management calls and its relationship to their use. 

. 2000;26(6):1003-1010. doi:10.1177/014572170002600613

24. Piette JD, Marinec N, Janda K, et al. Structured caregiver feedback enhances engagement and impact of mobile health support: a randomized trial in a lower-middle-income country. 

. 2016;22(4):261-268. doi:10.1089/tmj.2015.0099

25. Pfeiffer PN, Bohnert KM, Zivin K, et al. Mobile health monitoring to characterize depression symptom trajectories in primary care. 

. 2015;174:281-286. doi:10.1016/j.jad.2014.11.040

26. Piette JD, Aikens JE, Trivedi R, et al. Depression self-management assistance using automated telephonic assessments and social support. 

27. AbuDagga A, Resnick HE, Alwan M. Impact of blood pressure telemonitoring on hypertension outcomes: a literature review. 

. 2010;16(7):830-838. doi:10.1089/tmj.2010.0015

28. Maeder A, Poultney N, Morgan G, Lippiatt R. Patient compliance in home-based self-care telehealth projects. 

. 2015;21(8):439-442. doi:10.1177/1357633X15612382

29. O’Connor S, Hanlon P, O’Donnell CA, Garcia S, Glanville J, Mair FS. Understanding factors affecting patient and public engagement and recruitment to digital health interventions: a systematic review of qualitative studies. 

. 2016;16(1):120. doi:10.1186/s12911-016-0359-3

30. Shan R, Sarkar S, Martin SS. Digital health technology and mobile devices for the management of diabetes mellitus: state of the art. 

. 2019;62(6):877-887. doi:10.1007/s00125-019-4864-7

31. Piette JD, Marinec N, Gallegos-Cabriales EC, et al. Spanish-speaking patients’ engagement in interactive voice response (IVR) support calls for chronic disease self-management: data from three countries. 

. 2013;19(2):89-94. doi:10.1177/1357633x13476234

32. Steinman MA, Perry L, Perissinotto CM. Meeting the care needs of older adults isolated at home during the COVID-19 pandemic. 

. 2020;180(6):819-820. doi:10.1001/jamainternmed.2020.1661

Articles

Family support with medication management and recent urgent self-management concerns are 2 novel factors, among others, that predict completion of diabetes telehealth calls.

Patient and Supporter Factors Affecting Engagement With Diabetes Telehealth

: To examine willingness to participate in a pill-splitting program and the impact of pill splitting on patientsâ€™ adherence and lipid control. 

: A total of 200 patients who used statins and were candidates for a pill-splitting regimen were identified from a large university-based health plan. Sixty-three percent of study participants were female, 41% were nonwhite, and 94% had at least some college education. Patients were surveyed regarding their willingness to split pills, and 111 consented to participate in a 6-month trial in which half were randomized to receive a financial incentive to split pills: a 50% reduction in their per-refill copayment. Data on patientsâ€™ statin refills and lipid control were obtained from billing and medical records.

: Compared with patients unwilling to participate in the program, those agreeing to split pills were more likely to be female and white. After 6 months, most patients in the trial (89%) were willing to continue pill splitting for a 50% copayment reduction. Patients reported few problems with pill splitting and had no noticeable change in their adherence. The financial-incentive group and the control group did not differ significantly with respect to their low-density lipoprotein cholesterol levels after pill splitting: -2.0 mg/dL and -1.2 mg/dL, respectively. 

: Most patients indicated that at least a 50% copayment reduction would be required to enroll in a pill-splitting program after the study ended. However, in this relatively educated population, financial incentives did not influence patientsâ€™ adherence, satisfaction, or health outcomes.

(Am J Manag Care. 2007;13(part 1):298-304) 

A 50% reduction in patientsâ€™ per-refill copayment was used to promote adherence to a pill-splitting statin regimen in a randomized trial involving a relatively educated population.

Pill splitting did not adversely affect cholesterol levels, nor did it hinder compliance with medication therapies. 

During the 6-month trial, the financial incentive did not influence patientsâ€™ adherence, satisfaction, or health outcomes.

After the trial, 89% of patients were willing to continue pill splitting for

As medication costs continue to escalate, many institutions and managed care organizations are seeking to maximize the cost-effectiveness of individual pharmacotherapy. Pill-splitting programs have been implemented in various settings in an attempt to decrease medication costs.

 However, concerns have been raised that splitting could compromise patient safety due to uneven drug distribution in the split pills and loss of medication potency due to fragmentation.

 Although these concerns are legitimate for a fraction of the drugs available, for most medications, especially those with a long half-life, small variations due to fragmentation are not clinically important. In a randomized cross-over study, Rindone evaluated the impact of splitting lisinopril pills in 29 patients with hypertension; patients received a full pill for 2 weeks and a split pill for 2 weeks with no significant differences in systolic and diastolic blood pressures.

 Duncan et al evaluated the effects of pill splitting on low-density lipoprotein (LDL) and total cholesterol levels in a retrospective chart review of 125 patients receiving atorvastatin or simvastatin therapy who were switched to pill splitting. The results of that study demonstrated no statistically significant changes in LDL and total cholesterol levels.

Another study conducted by Gee et al found that pill splitting of statins had no negative effects on patients' cholesterol profile.

Concerns also have been raised about patient satisfaction and adherence with pill splitting.

 The few studies that examined these outcomes have reported variable results. McDevitt et al surveyed 94 patients with hypertension after 10 days of splitting their hydrochlorothiazide pills. Most of the subjects stated a preference for the commercially available lower-strength pills, and 77% reported a willingness to pay more for them.

 In contrast, Carr-Lopez et al reported that splitting lovastatin pills was easy to do and did not hinder medication adherence in the majority of patients.

 A study of fosinopril pill splitting reported no significant difference in medication adherence between patients who split pills and those who did not. In fact, the patients splitting fosinopril pills reported positive experiences.

 In the study by Gee et al, the majority of patients who split statin pills were satisfied and adherent to the new regimen.

Although these prior studies provide some insight, the majority of studies to date were retrospective, were based on very small samples, or evaluated effects after brief periods of splitting pills. Finally, it is unclear whether patients who split pills will be more satisfied and adherent to their regimens if they share in the financial benefit associated with this change.

The objectives of this study were (1) to evaluate the impact of pill splitting on clinical measures, patient satisfaction, and acceptance and (2) to determine whether a financial incentive would affect patients' decision to split pills.

In the current study, we addressed the previously mentioned issues by using a large survey sample with a nested randomized trial of financial incentives to promote adherence to a pill-splitting statin regimen. First, we recruited patients with hyperlipidemia who were candidates for pill splitting, and we assessed their perceptions about pill splitting. Second, survey respondents were invited to participate in a randomized trial in which they would receive a free pill cutter, would be encouraged to split their lipid-lowering medication, and would have a 50% chance of being randomized to a group who received a reduction in copayment for the split-pill regimen. We compared the characteristics of patients who did and did not consent to participate in the pill-splitting trial. Trial participants were randomized to either pill splitting only or pill splitting plus a 50% reduction in their per-refill copayment. Randomization was done using a blocked randomization and random-number generator with random block sizes of 2, 4, or 6.

After 6 months, we compared patients with and without the copayment reduction in terms of their satisfaction with pill splitting and willingness to continue splitting pills. Finally, we examined pre-post changes in the adherence and lipid levels of patients in the trial to determine whether patients who split pills had any evidence of altered lipid control.

The study was conducted among patients using cholesterol-lowering medications (statins) in a university-based healthcare system. Statin use provides an ideal context in which to evaluate the outcomes of an incentive-based splitting program because these treatments are common, and many patients use brand-name drugs that are expensive to third-party payers. Moreover, hyperlipidemia has a clear, regularly measured physiologic outcome that can be used to quantify any potential negative effects of splitting on patients' health status.

Patients who returned the signed consent form and baseline survey by mail indicated whether they would be willing to participate in a randomized trial of copayment reduction associated with pill splitting. All patients agreeing to participate in the trial received a patient education sheet on pill splitting and free pill cutters through the mail. Based on feedback from some of the patients, all patients received 2 different types of pill cutter in the mail to assess which type patients preferred. Their primary care physicians were contacted for a written prescription of the pill-splitting regimen or a verbal order enabling the study investigators to call in the prescription to the patient's pharmacy. The study investigators also called all participants to ensure that the pill cutter was being used properly and to answer questions within 1 month of enrollment in the trial.

Patients randomized to the copayment reduction group received a 50% discount on their copayment for refills of their statin medication, regardless of whether they filled those prescriptions at a university-based or a community-based pharmacy. Because participants differed in their initial copayments, the absolute dollar value of the decrease varied across copayment reduction patients. However, the majority of patients in this group had $7 or $5 savings a month depending on whether their prescription was filled at a retail pharmacy or by mail order, respectively. Patients randomized to the comparison group received no discount copayments for pill splitting. Patients were informed of their random assignment within 7 days of enrollment via mail or telephone.

All statistical calculations were performed using SAS for Windows 9.1 (SAS Institute Inc, Cary, NC). Using LDL data from the study by Duncan et al,

 we determined that a sample size of 96 subjects would be required to detect a between-group difference of 9 mg/dL in LDL concentrations with 90% power at 

 = .05 (2-sided test). We used χ2 tests and  cross-tabulations to identify variation across demographic groups in the survey sample's perceptions about pill splitting. We used similar statistics to compare the subsets of initial survey respondents who did and did not consent to participate in the financial-incentive trial and outcomes between the groups that did and did not receive the financial incentive. Pre-post comparisons of patients' adherence and lipid levels were based on either paired Wilcoxon signed rank tests for continuous data or McNemar's test for dichotomous variables. All analyses were conducted on an intent-to-treat basis.

Recruitment and Characteristics of the Study Sample

The majority of the 200 patients returning the baseline surveys were age 50 years or older, Caucasian, and had at least some college education. Nearly 80% of respondents rated their own health as very good or good, compared with only 12% who rated their health as excellent and only 8% who rated their health as fair or poor. Most respondents spent $50 or less per month on their prescriptions, but 5% reported spending more than $100 per month. Although only 11% reported missing 1 or more medication doses due to cost concerns, 86% of respondents reported being extremely or moderately concerned about the cost of their medications.

Most patients had a positive attitude about splitting pills, with 88% indicating a willingness to split pills in exchange for a reduced medication copayment and 84% willing to split pills to save money for their plan provider. When asked to identify how much money they would need to save per prescription for them to be willing to split pills, only 12% of respondents stated that pill splitting would not be worth any amount of monetary savings, and 59% stated that only a $5 to $10 savings per prescription would be needed for them to be willing to split pills. The remaining 57 respondents (29%) indicated that a $15 to $20 savings per prescription would be needed. 

Baseline characteristics of patients who agreed to participate in the pill-splitting trial were compared with those of respondents who declined trial participation. Patients who participated in the trial were more likely to be more than 50 years old, female, and taking more than 4 medications per month. Level of education did not correlate with patients' willingness to participate in the randomized phase of the study. There was no significant difference in health perceptions, amount of money spent on medications, compliance, or concerns about medication costs between trial participants and patients who only completed the baseline survey. However, there was a significant difference between the groups regarding their willingness to split pills to save money.

Overall, 89% of all trial participants expressed a willingness to continue splitting pills for a 50% copayment reduction until the degree of copayment reduction was determined for a future pill-splitting program, with a nonsignificantly larger proportion of those in the copayment reduction group expressing a willingness to continue pill splitting after the study (92.7% vs 85.2%) (

). Responses in the 2 trial groups were generally similar with respect to the perceived effort of splitting pills, and when asked whether they agreed with the statement that the “effort to split pills is no big deal,â€ approximately 80% of trial participants either “strongly agreedâ€ or “agreed.â€ Patients were asked to select how much of a copayment reduction would be necessary for them to participate in a pill-splitting program after the end of the study. Of the 3 possible response categories, 73% reported that they would require a 50% reduction in copayment, 26 (24%) stated that they would need their copayment waived, and 3 (3%) reported that splitting would not be worth the effort regardless of savings. Overall, there were few differences in the required copayment for splitting pills between the group who received a copayment reduction and group who did not.

Patients in the 2 groups had few significant differences in their willingness to continue pill splitting after the trial. Also, individuals willing to continue splitting pills tended to take more medications per month (27% took 7 or more medications per month vs only 8% of those who were not willing to continue splitting pills). Patients willing to split pills also spent more per month out of pocket on their medications, although these differences did not reach statistical significance.

Impacts of Pill Splitting on Medication Adherence Participants in the trial strongly indicated that splitting pills did not negatively affect their adherence. More than 90% either agreed or strongly agreed with the statement that “pill splitting has not affected my willingness to take my medication,â€ and fewer than 5% disagreed or strongly disagreed. Similarly, 90% disagreed or strongly disagreed with the statement that they had “missed more medication doses because of pill splitting.â€ Only 5% reported having 3 or more occasions on which they had to discard pills due to splitting-related problems, compared with 92 (84%) who reported zero such occasions and 12 (11%) who reported only 1-2 such occasions. Self-reported adherence rates were similar for patients with and without the copayment reduction. 

A total of 94 trial participants had at least 2 prescription fills identified during the follow-up period, allowing us to calculate MPRs. The mean Â± SD MPR in the 6 months prior to tablet splitting was 94% Â± 17%. The mean Â± SD MPR during the pill-splitting period was virtually unchanged at 94% Â± 14%.

Overall, lipid profiles did not change significantly over the course of the pill-splitting period in either patients randomized to receive the copayment reduction or those with no copayment reduction (

This study is the first prospective randomized trial evaluating the impact of pill-splitting financial incentives on patients' adherence, satisfaction, and health outcomes. It is also unique because the population includes more women and because pretrial and posttrial surveys were used in conjunction with lipid data. With respect to the impact of pill splitting on patients' adherence and health status, our results were very similar to those of other pill-splitting studies. Overall patient satisfaction was high, patients' medication adherence did not change once splitting was initiated, and lipid levels were not negatively affected.

This study provides insight into voluntary pill splitting accompanied by sharing of insurance plan cost savings with patients. The copayment incentive did not influence patient satisfaction, adherence, or cholesterol profiles during the study. Although many participants in the trial reported that they would require a financial incentive to continue splitting pills, the impact of the incentive in the randomized study suggests that in fact such incentives may not influence their willingness to adhere to a split-pill regimen.

Rather than financial incentives, this study identified specific characteristics of a split-pill regimen that can affect patients' acceptance. One of the main reported reasons for dissatisfaction with splitting was difficulty in using the pill cutter for some patients (eg, failure to yield evenly split pills, crushing the tablet). Recognizing the importance of using the proper pill cutter, we provided 2 different types of pill cutters: a small blue splitter without rubber padding and a clear splitter with both rubber padding and a wider cutting area to fit larger pills. Most of the patients preferred the clear splitter, especially those patients taking larger-sized pills. Another reported reason for dissatisfaction with splitting was difficulty in swallowing split pills due to rough edges and bitter taste (reported by 3 patients).

One of the limitations of our study is that it was conducted at a single university-based health center. It remains to be seen whether pill splitting will be well received and accepted across our whole healthcare system or within health systems treating a more socioeconomically and racially diverse patient population. Compared with some previous pill-splitting trials, our study population was younger and highly educated. Although income data were not collected, such patients may have had relatively high incomes and they may have been less sensitive to the small cost savings associated with pill splitting. Also, we did not distinguish between patients receiving statins for primary versus secondary prevention in our data collection. We found that patients less likely to be interested in pill splitting included those who had more formal education, spent less on medications, and took fewer medications. In conjunction with other studies, this result suggests that patients receiving Medicaid or other financially pressed pharmacy insurance programs may be more willing to split pills if the program effectively addresses patients' concerns about using the splitter. Another limitation of this study was its short duration. Our prescription plan allowed for a 3-month supply per refill, providing only 2 opportunities to observe compliance. This may have biased the MPR estimates to reflect overall optimistic adherence levels compared with what would have been observed given a longer follow-up period.

The financial benefits of moving to a split-pill regimen can be enormous. Since this study was completed, we have implemented a pill-splitting program for all university employees through the university pharmacy benefit plan. All patients will be offered a 50% copayment reduction for splitting atorvastatin, pravastatin, and simvastatin, with a free pill cutter provided once a year. Without much marketing effort, 5.3% of statin users were splitting pills within less than 3 months of implementation of the pill-splitting program, yielding $33 000 savings. Extrapolated annual cost savings of this program would be $1.3 million if 45% of current statin users started to split pills.

When pill-splitting plans are implemented, only the prescription plan benefits financially unless there is some sharing of cost savings with patients. Although in our study sharing cost savings with patients did not significantly impact their willingness to participate, most patients said that they would expect some financial incentive in order to continue splitting pills. Because there are a variety of patient and product factors involved, voluntary pill-splitting programs seem to be more optimal when some of the financial benefits are shared with patients. This strategy should increase patient participation in a nonmandatory pill-splitting program. In any program, pill splitting should be implemented with consideration of which pills can be split without compromising clinical outcomes, as well as recognizing that this practice is safe only in patients who are able to cognitively and physically split pills. 

The current study found that many patients using lipid-lowering therapies, particularly those using a larger number of drugs, were willing to participate in a pill-splitting program. Although patients reported that financial incentives would increase their willingness to split pills, our randomized trial did not indicate that such incentives had influenced patients' adherence, satisfaction with splitting, or lipid levels.

John Piette, PhD, is a VA Research Career Scientist. Michele Heisler, MD, MPA, is a VA Career Development Awardee. Jennifer Kim, PharmD, and Diane Domas, PharmD, were University of Michigan research assistants for this study. We wish to thank Dawn M. Parsons, RPH, MBA, for assisting in the system set up of PBM. 

: From the College of Pharmacy (HMC, JGS, DSS) and the Department of Internal Medicine (MH, CJS, JDP), University of Michigan, and the University of Michigan Hospitals and Health Centers (HMC, JGS, CJS), Ann Arbor, Mich; the Center for Practice Management and Outcomes Research, VA Ann Arbor Health Care System (MH, JDP); and the Michigan Diabetes Research and Training Center, Ann Arbor (JDP).

 The University of Michigan Employee Benefits Office provided funding for this study. The principal investigator had full access to all of the data in the study

and takes responsibility for the integrity of the data and the accuracy of the data analysis. 

: Hae Mi Choe, PharmD, East Ann Arbor Health Center Pharmacy, 4260 Plymouth Rd, Ann Arbor, MI 48109-2704. E-mail: 

: The authors (HMC, JGS, DSS, MH, CJS, JDP) report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter discussed in this manuscript. 

: Concept and design (HMC, JGS, DSS, MH, CJS, JDP); acquisition of data (HMC, JDP); analysis and interpretation of data (HMC, JGS, DSS, MH, JDP); drafting of the manuscript (HMC, JGS, DSS, MH, CJS, JDP); critical revision of the manuscript for important intellectual content (HMC, JGS, MH, CJS, JDP); statistical analysis (JGS, DSS, JDP); provision of study materials or patients (HMC); obtaining funding (HMC, CJS); administrative, technical, or logistic support (HMC, JGS); and supervision (HMC).

. Relationship between pill splitting and compliance, drug acquisition cost, and patient acceptance. 

. Effects of a tablet-splitting program in patients taking HMG-CoA reductase inhibitors: analysis of clinical effects, patient satisfaction, compliance, and cost avoidance. 

. The practice of splitting tabletsâ€” cost and therapeutic aspects. 

Impact of Patient Financial Incentives on Participation and Outcomes in a Statin Pill-splitting Program

Our understanding of diabetes and its managementis evolving rapidly based on informationfrom researchers and the world of healthcaredelivery about "what works" to improve outcomes.This theme issue of the 

includes studies thatcan inform efforts to improve diabetes care on manylevels, including more effective clinical decision-making,improvements in health systems that promotetreatment quality, and health policies that encouragethe best disease management practices.

Despite years of research on diabetes managementand quality improvement, these new studies are worthwhilebecause, unfortunately, diabetes treatmentprocesses and outcomes remain disappointing in manyhealthcare systems. The study by Godley and colleagues

John D. Piette, PhD

Articles

Patient and Supporter Factors Affecting Engagement With Diabetes Telehealth

Impact of Patient Financial Incentives on Participation and Outcomes in a Statin Pill-splitting Program

The Future of Diabetes Disease Management: Integrating Lessons Learned from Clinical, Health Services, and Policy Research

Problems Due to Medication Costs Among VA and Non-VA Patients With Chronic Illnesses

Addressing the Needs of Patients With Multiple Chronic Illnesses

Interactive Voice Response Systems in the Diagnosis and Management of Chronic Disease