Jaan Sidorov, MD, FACP, CMCE, Robert Shull, PHD, Janet Tomcavage, RN, MSN, CDE, Sabrina Girolami, RN, BSN, Ronald Harris, MD, FACE, Nadine Lawton, RN
From: “Diabetes Care. 2002;25(4)”
Objective. Little is known about the impact of disease management programs on medical costs for patients with diabetes. This study compared health care costs for patients who fulfilled health employer data and information set (HEDIS) criteria for diabetes and were in a health maintenance organization (HMO)-sponsored disease management program with costs for those not in disease management.
Research Design And Methods. We retrospectively examined paid health care claims and other measures of health care use over 2 years among 6,799 continuously enrolled Geisinger Health Plan patients who fulfilled HEDIS criteria for diabetes. Two groups were compared: those who were enrolled in an opt-in disease management program and those who were not enrolled. We also compared HEDIS data on HbA1c testing, percent not in control, lipid testing, diabetic eye screening, and kidney disease screening. All HEDIS measures were based on a hybrid method of claims and chart audits, except for percent not in control, which was based on chart audits only.
Results. Of 6,799 patients fulfilling HEDIS criteria for the diagnosis of diabetes, 3,118 (45.9%) patients were enrolled in a disease management program (program), and 3,681 (54.1%) were not enrolled (nonprogram). Both groups had similar male-to-female ratios, and the program patients were 1.4 years younger than the nonprogram patients. Per member per month paid claims averaged $394.62 for program patients compared with $502.48 for nonprogram patients (P < 0.05). This difference was accompanied by lower inpatient health care use in program patients (mean of 0.12 admissions per patient per year and 0.56 inpatient days per patient per year) than in nonprogram patients (0.16 and 0.98, P < 0.05 for both measures). Program patients experienced fewer emergency room visits (0.49 per member per year) than nonprogram patients (0.56) but had a higher number of primary care visits (8.36 vs. 7.78, P < 0.05 for both measures). Except for emergency room visits, these differences remained statistically significant after controlling for age, sex, HMO enrollment duration, presence of a pharmacy benefit, and insurance type. Program patients also achieved higher HEDIS scores for HbA1c testing as well as for lipid, eye, and kidney screenings (96.6, 91.1, 79.1, and 68.5% among program patients versus 83.8, 77.6, 64.9, and 39.3% among nonprogram patients, P < 0.05 for all measures). Among 1,074 patients with HbA1c levels measured in a HEDIS chart audit, 35 of 526 (6.7%) program patients had a level >9.5%, as compared with 79 of 548 (14.4%) nonprogram patients.
Conclusions. In this HMO, an opt-in disease management program appeared to be associated with a significant reduction in health care costs and other measures of health care use. There was also a simultaneous improvement in HEDIS measures of quality care. These data suggest that disease management may result in savings for sponsored managed care organizations and that improvements in HEDIS measures are not necessarily associated with increased medical costs.
Diabetes is associated with significant health care costs. It has been estimated to affect 16 million Americans, with $44 billion a year in direct medical and treatment costs. Although people with diabetes account for only 3.8% of the U.S. population, this disease accounts for 5.8% of all personal health care expenditures in the U.S. .
The cost of diabetes care for managed care organizations (MCOs) is also substantial. MCO enrollees with diabetes have higher rates of cardiovascular, eye, lower-extremity, and renal disease compared with those without diabetes [2–4]. Several studies have conclusively demonstrated that complications from diabetes can be reduced by aggressive glycemic control [5–10]. As a result, many MCOs have sponsored initiatives to improve glycemic control among their members in the belief that this will reduce the rate of diabetes complications and associated health care costs.
Improving health outcomes and lowering use and costs underlie the strategy of disease management. Disease management is defined as any multifaceted program devoted to the care of populations characterized by the presence of a chronic disease. Disease management programs are usually financed with a fixed percentage of the insurance premium. If complication rates are lowered, the lower use and associated savings can result in profit for the sponsoring organization. Characteristics of disease management programs typically include disease staging, promotion of clinical guidelines, patient education that promotes self-management, aggressive screening for complications, and early and appropriate specialty referral [11–17].
Little is known about the impact of disease management programs on health care costs, quality of care, and complication rates among patients with diabetes. Disease management programs for diabetes vary in scope and content, and a significant number are offered by independent companies under proprietary circumstances [17–19]. It is also unclear whether disease management can result in short-term savings because the consequences of poor glycemic control occur over many years [20–24].
In this study, we describe the short-term medical cost savings associated with a health maintenance organization (HMO)-sponsored disease management program by comparing the claims of enrollees who fulfilled health employer data and information set (HEDIS) criteria and were in disease management with those not in disease management.
Geisinger Health Plan (GHP) is a federally qualified, not-for-profit group model HMO with 295,000 enrollees in 41 counties in northeastern and central Pennsylvania. It is part of the Geisinger System, which also supports a multispecialty group practice clinic consisting of 587 physicians located in 64 clinic sites as well as two closed-staff hospitals. GHP also independently contracts with a network of 4,192 providers and 57 hospitals. Several types of managed care insurance are offered by GHP, including commercial, Medicare risk, small business, group, individual, and third-party administration (TPA). Seventy-three percent of enrollees use one of a series of pharmacy benefit packages that can be purchased as a separate rider with a variety of patient co-pay options. Glucose monitors and strips are considered durable medical equipment and are covered unless specifically excluded, as negotiated under a TPA arrangement. GHP is fully accredited by the National Committee on Quality Assurance. As part of the accreditation process, GHP conducts yearly measures of the quality of diabetes care using HEDIS criteria, which is a set of performance measures obtained using a proscribed methodology designed to enable purchasers to reliably compare the performance of different managed health care plans .
On 1 April 1997, GHP’s disease management department began to recruit patients for diabetes disease management. At the time of this study, our department used a network of 51 primary care nurse educators and case managers. These registered nurses provide patient education and case management services in all physician clinics that contract with GHP for primary care services. Depending on member enrollment and geographic proximity, each nurse is responsible for 1–15 primary care sites. Each nurse is trained in diabetes patient education as well as tobacco cessation, congestive heart failure, hypertension, chronic obstructive pulmonary disease, and asthma. In fiscal year 2000, the total cost of this program, including capital, totaled just over $4.2 million, with 4,262 continuously and currently enrolled patients who entered diabetes disease management. This consists of ~43% of all patients ever seen by the disease management nurses .
Patient education by the nurses at each primary care clinic is provided one on one or in group settings by appointment. Each nurse encounter is documented in the patient’s medical record for physician review and co-signature. Nurses are allowed to use judgment in accommodating local physician preferences and practice styles. There is no charge rendered for the nurse education and no net financial gain or loss for the primary care site. Each nurse is also responsible for baseline and ongoing collection of data from the patient or from the medical record for later entry in database registries. Use of this approach to achieve outcomes in the areas of tobacco cessation, living wills, and diabetes care has been previously reported [27–29].
Description of diabetes disease management
A detailed description of the GHP diabetes disease management program has been described elsewhere . Briefly, this is a package of interventions, given over 1 year, consisting of promotion of diabetes clinical guidelines by the nurses in their day-to-day interactions with primary care physicians and patients, HMO-sponsored continuing medical education sessions for primary care providers, early and appropriate specialty clinic referral, and primary care site-based patient education and case management by the HMO nurses. Patients must voluntarily opt in to participate. To aid recruitment, nurses can arrange the one-time provision of a glucose meter and 100 glucose meter strips at no cost, using clinical criteria from the diabetes guidelines. Additional glucose meter strips are available for monthly co-payment, ranging from $8 to $15. Any patient with diabetes may self- refer or be referred by their physician. Depending on patient and physician preference, baseline HbA1c measurement, and the presence of any diabetic complications, all patients are seen one to four times by the nurse from the date of referral. All participants are educated about the appropriate use of a glucose meter, the role of diet and exercise, the importance of HbA1c testing, medication management, the management of hypoglycemia, and teaming closely with physicians in the use of staged diabetes management clinical guidelines  to achieve optimum blood glucose control.
Analysis of savings in diabetes disease management
GHP enrollees eligible for HEDIS analyses at the time of this study totaled 172,015 commercial HMO, 36,456 Medicare risk, and 47,004 patients with “point-of-service” insurance. Of 255,475 HMO enrollees, 6,799 (2.7%) fulfilled HEDIS criteria for the presence of diabetes. Of this latter group, 3,118 (45.8%) had been seen at least once by a GHP nurse since the program began in April 1997. HEDIS-specific data on all patients were obtained by a separate group of nurses (in the case of chart reviews) or data analysts (in the case of claims extracts) devoted to measuring quality improvement outside of the disease management program. Personnel responsible for collecting or reporting HEDIS data were unaware of those patients that were in disease management at the time of their review.
All GHP members with commercial, including point-of-service, or Medicare risk insurance fulfilling HEDIS criteria for diabetes during the 2-year period from 1 July 1999 to 30 June 2001 had all enrollment data and submitted claims for health or medical care downloaded from the HMO claims database and entered into SAS version 8.0. The criteria used to identify members who fulfill HEDIS criteria are described elsewhere . Pharmacy claims were not included in this analysis. Unique member identification numbers were sorted into those who had seen an HMO disease management nurse at least once for diabetes education (program patients) and those not entered into disease management (nonprogram patients). Mean total claims paid per member per month, mean admissions per patient per year, mean number of inpatient days per patient per year, and mean number of emergency room and primary care office visits were compared in the two groups. We also compared HEDIS scores for HbA1c testing, percent not in control, diabetic eye screening, and kidney disease screening in the two groups. Duration or specific type of diabetes is not included in any HEDIS measure, and this information is not included in this analysis. 2tests were used to examine the significance of any observed differences in tests of proportion. Student’s t tests were used to examine the statistical significance of any observed differences in tests of continuous data. Multiple linear regression and the resulting F statistic was used to control for age, sex, presence of pharmacy benefit, HMO enrollment duration, and insurance type to more precisely describe the significance of continuous data.
There were 6,799 continuously enrolled patients who fulfilled diabetes HEDIS criteria during the 2 years of this study. A total of 3,118 (45.9%) subjects had enrolled in disease management and were managed by 51 disease management nurses, as compared with 3,681 (54.1%) subjects who were not in disease management. The average number of visits with an HMO nurse was 3.63. A total of 419 (13.43%) patients visited a nurse one time, and 2,699 (86.57%) visited a nurse two or more times. Table 1compares the male-to-female ratio, age, HMO enrollment duration, presence of a pharmacy benefit plan, and insurance type (commercial versus Medicare risk) between the study subjects fulfilling HEDIS criteria who were in disease management (program) and those fulfilling HEDIS criteria who were not in disease management (nonprogram). Of the demographic variables, sex was not significantly different between the two groups (P > 0.05), but program patients were on average 1.4 years younger (P < 0.05), had longer enrollment duration in the HMO, were more likely to have a pharmacy benefit plan, and were more likely to have commercial insurance (P < 0.05 for all four measures).
During the 2-year period of study, program patients experienced $394.62 per member per month in mean total paid claims, as compared with $502.48 for those not in disease management (P < 0.05, Student’s t test). This difference was accompanied by lower inpatient use among program patients, who experienced a mean of 0.12 admissions per patient per year and 0.56 inpatient days per patient per year, as compared with nonprogram patients, who had 0.16 admissions and 0.98 inpatient days per patient per year (P < 0.05 for both measures.). The mean number of emergency room visits was 0.49 per patient for program patients compared with 0.56 among nonprogram patients (P < 0.05). In contrast to emergency room use, program patients experienced a higher mean number of primary care office visits (8.4) per patient per year compared with nonprogram patients (7.8). When these data were compared while statistically controlling for age, sex, enrollment duration, presence of a pharmacy benefit, and insurance type, all measures of use, except for emergency room visits, remained statistically significant ( Table 1 ). When enrollees in commercial and Medicare risk insurance lines were compared separately, statistically significant lower mean paid claims per member per month among the program patients, as compared with nonprogram patients, persisted ($302.19 vs. $527.96 and $424.00 vs. $500.37, respectively, P < 0.05 for all measures).
Program patients also experienced favorable HEDIS scores compared with nonprogram patients. HbA1c testing as well as lipid, eye, and kidney screening were 96.6, 91.1, 79.1, and 68.5%, respectively, among program patients compared with 83.8, 77.6, 64.9, and 39.3%, respectively, among nonprogram patients. All observed differences were statistically significant (P< 0.05). A total of 1,074 patient charts (526 program and 548 nonprogram patients) were reviewed for determination of HbA1c under control. Thirty-five (6.7%) program patients had HbA1c >9.5% compared with 79 (14.4%) nonprogram patients ( Table 1 ).
These retrospective data demonstrate that participants in a managed care-sponsored diabetes disease management program experienced lower overall paid insurance claims for health care compared with those not in disease management. This difference was not only statistically significant but substantial, amounting to $104.86 per member per month or $ 1,294.32 per year. For the 3,118 continuously enrolled patients included in this analysis, this amounts to a total of $4,035,689.70 per year in fewer claims paid compared with nonprogram patients. Lower claims for program patients were present in both commercial and Medicare risk insurance. As noted above, the total budget, including capital for all disease management programs in this HMO, was ~$4.2 million per year. Because ~43% of all patients seen in disease management had diabetes, we believe the estimated allocated cost of ~$1.81 million for diabetes disease management contrasts favorably with the $4,035,689.70 in fewer claims for the patients included in this analysis. We found that much of the observed savings were accompanied by comparatively lower measures of inpatient use, with fewer admissions and fewer inpatient days. These findings persisted after we statistically controlled for factors that could alter health care use, such as age, sex, duration of enrollment in the HMO, presence of a pharmacy benefit, and type of insurance. Because all insurance claims for each year of the study were recorded among the HMO enrollees we examined, it is unlikely that the savings were underestimated .
Our findings also add to the weight of evidence linking diabetes disease management to health care use and glycemic control. We found that patients in disease management not only experienced lower charges but also had significantly higher measures in the key diabetes HEDIS measures. Although our data do not support the assertion that increased quality causes lower health care costs, we did find it is possible to achieve both at the same time. This association between cost and glycemic control has been previously described. Davies et al.  examined the effectiveness of nurse-based diabetes education and found less inpatient use was associated with better glycemic control. Menzin et al.  also linked insurance claims and mean HbA1c levels among 2,394 patients with diabetes in the Fallon Clinic Health Plan. As in this study, the economic impact of blood glucose control was apparent within a relatively short period of time and was also manifested by less inpatient use. Gilmer et al.  and Wagner et al.  also found hospitalizations and overall health care costs in a managed care setting to be positively associated with elevated HbA1c levels. Others outside of managed care have shown that in randomized clinical trials, achieving a lower HbA1c is associated with fewer complications and lower health care costs [36,37].
These data also support the findings of other researchers who have shown that nurses can champion clinical guidelines and provide diabetes education to achieve significant improvements in blood glucose control [38–45]. This approach compares favorably with usual primary care, in which up to 40% of patients with diabetes do not have a measurement of their HbA1c. Aubert et al.  found that in a randomized clinical trial, nurse managers can achieve significant improvements in blood glucose among primary care patients. As in this program, these nurses relied on staged diabetes management guidelines, which also have been shown to result in better glycemic control .
To our knowledge, this is the first report linking HEDIS and use. HEDIS theoretically enables purchasers to compare quality among competing MCOs. Purchasers also use other considerations when choosing an MCO, such as premium amount, network size, and financial stability. Despite widespread use of HEDIS, managed care has been criticized for failing to convince purchasers to rate quality of care over other factors in purchasing decisions . Our data suggest that patient education, clinical guidelines with provider teaming, and financial performance need not be mutually exclusive.
The growth of independent disease management companies, financed through a percentage of the insurance premium, is further evidence of a widespread belief that this strategy can achieve bottom-line savings. Reports of their success across a variety of managed care settings, in lowering use or improving outcome measures, also stress the effectiveness of clinical guidelines and team-based care, which promotes self-management [50–54].
Our findings may be biased. For example, greater willingness to cooperate with treatment recommendations, better health practices, or more interest in use of a glucose meter among patients who also agreed to opt in could explain the differences in use rather than disease management per se. In addition, because physicians had referred an unmeasured fraction of program patients, some of the differences in use could have been the result of differences in physician behavior outside of the disease management program. However, this program recruited just under one-half of all patients fulfilling HEDIS criteria for diabetes from the same network of primary care sites that cared for patients not in disease management. We also statistically controlled for known patient variables that could have accounted for the observed outcome differences. Because this disease management program was available to all HMO members with diabetes, close to one-half of eligible patients used it. We statistically controlled for known confounding patient variables, and we believe the impact of other unmeasured patient factors could have caused nonrandom selection and fewer claims, reducing better outcomes in the disease management group. Regardless of potential patient selection bias, our simultaneous demonstration of improved clinical outcomes and lower use has important implications for health care organizations struggling to reconcile cost and quality.
Our data were also limited by restricting the claims analysis to overall paid charges. Although we found evidence of decreased inpatient use (manifested by fewer admissions and fewer inpatient days in program patients) and increased primary care office visits, we were unable to more fully characterize the savings. Although insurance claims are linked to diagnoses by ICD-9 code, we have anecdotally found that practice patterns and reimbursement issues significantly influence code selection, thus limiting our analysis. We were also unable to determine whether the HMO education nurses influenced health care use by redirecting their patients away from more costly services. We also caution that fewer insurance claims for health care do not necessarily mean lower health care costs, especially for patients who may experience significant out-of-pocket expenses. Our data are also limited by the lack of information concerning the use or cost of pharmaceuticals, which could also be responsible for changes in use. Our population resides in a largely rural setting, which may also limit the generalizability of our findings. Finally, this disease management program consisted of several interventions that in turn were adapted to accommodate local physician practice styles. Determining the source of short-term savings in disease management using methodologies that can prospectively and precisely define the relative contribution of each of the interventions typically used in multifaceted disease management programs is an area ripe for further research.
These issues can only be addressed through random selection and assignment of patients in a clinical trial using predefined clinical and financial criteria. Pending more research in this area, however, our data may demonstrate that disease management can simultaneously benefit participants and MCOs, with lower health care use, significant savings, and higher health care quality.
Abbreviations: GHP, Geisinger Health Plan • HEDIS, health employer data and information set • HMO, health maintenance organization • MCO, managed care organization • TPA, third-party administration
The Development of Prediabetes and Diabetes
Your body will try to keep up with demands, and can be successful for a long time. As your insulin resistance increases, your pancreas can produce increasing amounts of insulin to try to keep blood sugar levels in normal ranges.
If this continues, though, the system will eventually break down. At some point, your pancreas will be unable to produce enough insulin to keep up with demand. This can be because your cells develop a high level of insulin resistance, and/or because your pancreatic beta cells, which produce insulin, become exhausted.
This point, when insulin is no longer sufficient, is the point when your blood sugar levels begin to rise. You can get a blood test, and your doctor can diagnose you with prediabetes or diabetes based on the result. I
Blood Sugar Levels in Prediabetes
There are three tests you can get to check for prediabetes. They are all related to your blood sugar. Prediabetes values are higher than the values that are considered normal, but lower than the values in diabetes.
|Test||What It Measures||How It Works||Prediabetes Value|
Glycated hemoglobin (A1C)
Longer-term (3-month) estimate of blood sugar values
Simple blood test – does not need to be fasting
5.7 to 6.4%
Fasting plasma glucose (FPG)
The amount of glucose in your bloodstream
Blood test after an overnight fast (at least 8 hours)
Oral glucose tolerance test (OGTT)
How well your body metabolizes glucose (sugar)
Blood test 1 hour after you drink a solution with 75 grams of glucose
A new study reveals that artificial intelligence could be a useful tool to help patients prevent Type 2 diabetes. The study, published in the Journal of Medical Internet Research, showed that patients at risk of Type 2 diabetes who used the Lark Weight Loss Health Coach AI, dropped 2.38 percent of their baseline weight and increase the percentage of healthful meals they ate by 31 percent.
“I’m really excited,” Lark CEO and cofounder Julia Hu told MobiHealthNews. “I think stepping back and looking at pre-diabetes, it really is such a crushing and chronic condition. Eighty-six million people in the US have pre-diabetes and it costs the health industry billions of dollars a year.”
The longitudinal observational study was a partnership between six primary care offices in Nevada and southern California and Lark Technologies. The study examined 239 overweight and obese patients at risk of Type 2 diabetes. Participants were offered the app free of charge by their primary care physician. No further physician support was given to the patients, according to the study. The retrospective study looked at users from July 2016 to January 2017.
The app is personalized to meet patients’ needs and goals. The user can enter their age, gender, weight, height, and goals.
“This is really AI coaching and mimicking healthcare at its best, but the cool thing about AI is that it is infinitely scalable,” said Hu.
The study, noted that one of its limitations was the fact that it did not have a control group to directly compare the results to. In addition, the study was observational and not experimental, which meant it was not able to determine causality.
Lark’s HCAI has been on the market since 2015 and currently has about a million users, according to Hu. It provides weight loss coaching through modules on an array of topics in unlimited text-based counseling sessions. The modules take about 16 weeks depending on whether users miss a week of the classes or decide to rewatch a module.
According to the study, the design takes a holistic approach to weight loss and management. It allows users to report feelings, such as guilt, and then gives the user encouraging advice based on that situation.
“I grew up with a chronic disease all my life,” said Hu. “I had an amazing care team, my pediatrician and my dad. They were there for me not just medically but emotionally … so that made a huge impact on me, people struggling with chronic conditions could really use a friend and someone who is there for them when they need it. We wanted to scale that sense of compassion and relationship.”
The company has a healthcare committee of 15 members, including several Harvard and Stanford faculty members that specialize in psychology of some kind.
Lark Technologies debuted in 2011 and was originally focused on sleep technologies, but have pivoted, or as Hu says, evolved. Now there are no longer field devices but the technologies sit on top of 70 different health monitors.
“The idea was, can we create an AI coach that coaches on several different aspects of your health,” said Hu. “We started with sleep then, went to exercise, then went to nutrition and that helped us really build for chronic conditions. They are diseases that require lifestyle changes, behavioral health change, and self efficacy.”
Currently, Lark has four products on the marketing including a wellness, diabetes prevention, diabetes management, and hypertension management platforms.
“This study demonstrates AI’s potential to provide compassionate care that is associated with weight loss, increased healthy lifestyle behaviors, and user trust that can reduce diabetes risk,” the study reads.