- Email: [email protected]
Effects and Factors Related to Adherence to A Diabetes Pay-for-Performance Program: Analyses of a National Health Insurance Claims Database
JAMDA xxx (2016) 1e7
JAMDA journal homepage: www.jamda.com
Original Study
Effects and Factors Related to Adherence to A Diabetes Pay-for-Performance Program: Analyses of a National Health Insurance Claims Database Mei-Ju Chi PhD a, b, Kuei-Ru Chou PhD c, d, e, Dee Pei MD, PhD f, g, Jawl-Shan Hwang MD h, i, Laurie Quinn PhD j, Min-Huey Chung PhD c, Yuan-Mei Liao PhD c, * a
School of Gerontology Health Management, College of Nursing, Taipei Medical University, Taipei City, Taiwan Master Program in Long-term Care, College of Nursing, Taipei Medical University, Taipei City, Taiwan School of Nursing, College of Nursing, Taipei Medical University, Taipei City, Taiwan d Psychiatric Research Center, Taipei Medical University Hospital, Taipei City, Taiwan e Department of Nursing, Taipei Medical University-Shuang Ho Hospital, New Taipei City, Taiwan f College of Medicine, Fu Jen Catholic University, New Taipei City, Taiwan g Department of Endocrinology, Cardinal Tien Hospital, New Taipei City, Taiwan h Division of Endocrinology and Metabolism, Department of Internal Medicine, Chang Gung Memorial Hospital-Linkou, Taoyuan City, Taiwan i College of Medicine, Chang Gung University, Taoyuan City, Taiwan j Department of Biobehavioral Health Science, College of Nursing, University of Illinois at Chicago, Chicago, IL b c
a b s t r a c t Keywords: adherence diabetes healthcare expense healthcare utilization pay-for-performance
Objectives: To compare the effects of a diabetes pay-for-performance (P4P) program on diabetes-related/ nondiabetes-related healthcare utilization/expenses between participants who adhered to the program and those who did not, and explore factors related to program adherence. Design: A secondary data analysis with a natural experimental design. Setting: Taiwan’s National Health Insurance claims database (2001e2011) of newly diagnosed patients with diabetes in 2001 was used for the analyses. Participants: The database under analyses contained 119,970 patients who were newly diagnosed with diabetes in 2001. Longitudinal data from 2001 to 2011 were obtained. A sample of 5592 patients who were enrolled in the diabetes P4P program during 2003e2006 was identified. After a 3-year follow-up of the enrolled patients, 2647 (47.3%) of them adhered to the program. To minimize the differences between the characteristics of the patients who adhered to the program and those who did not, propensity score matching was adopted. A total of 5294 patients (adherence: 2647 vs nonadherence: 2647) were included for analyses. Measurements: We estimated utilization/expenses of healthcare services for both groups at 6 time points and applied t tests to test each utilization and expense of healthcare services between the 2 groups. A repeated-measures analysis of variance was applied to examine changes in the annual diabetes-related healthcare service expenses and total annual expenses by group. Logistic regression models were used to examine factors related to program adherence. Covariates included participant age, gender, diabetesrelated complications, Charlson Comorbidity Index, Continuity of Care Index, time since diagnosis of diabetes, hospitalization in the previous year, and location receiving healthcare services. Results: Total annual healthcare expenses spent by the adherence group were significantly lower than those of the nonadherence group. Gender, continuity of care, time since diagnosis of diabetes, hospitalizations in the previous year, and location receiving healthcare services were factors related to program adherence. Conclusions: Long-term, beneficial effects of the diabetes P4P program might have been present if patients had adhered to the program. Interventions and strategies which could improve program adherence and continuity of care are suggested to achieve optimal disease control and clinical outcomes. Ó 2016 AMDA e The Society for Post-Acute and Long-Term Care Medicine.
This work was supported by a grant from the College of Nursing, Taipei Medical University. The authors declare no conflicts of interest.
http://dx.doi.org/10.1016/j.jamda.2016.02.033 1525-8610/Ó 2016 AMDA e The Society for Post-Acute and Long-Term Care Medicine.
* Address correspondence to Yuan-Mei Liao, PhD, School of Nursing, College of Nursing, Taipei Medical University, 250 Wu-Xing St, Xinyi District, Taipei City, Taiwan 110. E-mail address: [email protected] (Y.-M. Liao).
2
M.-J. Chi et al. / JAMDA xxx (2016) 1e7
Diabetes affects more than 300 million individuals globally.1 Diabetes is one of the top 10 causes of death worldwide2; it is Taiwan’s fourth or fifth leading cause of death.3 The population coverage rate of Taiwan’s National Health Insurance (NHI) has reached 99.9%.4 Analyses of Taiwan’s NHI claims database revealed that the prevalence of diabetes increased 35% from 2000 to 2009, and the total population with diabetes increased more than 70%.5 Health spending on diabetes accounted for 10.8% of the worldwide total health expenditure.6 In Taiwan, the total healthcare cost for individuals with diabetes was approximately 2.8 times of the cost spent by the age-matched and sex-matched individuals without diabetes, and the total expenditure of diabetes to society is about US$2.96 billion.7,8 In Taiwan, a diabetes shared-care program was initiated in 1996; it is a comprehensive, integrated approach to provide healthcare and reimbursement based on a disease’s natural course.9 Interest in payfor-performance (P4P) as a strategy to stimulate the delivery of quality care is increasing; P4P provides financial rewards to healthcare providers for quality improvements.10 To enhance the operation of Taiwan’s diabetes shared-care program, a diabetes P4P program was implemented in 2001. In Taiwan’s diabetes P4P program, financial incentives include additional physician fees for providing comprehensive care, fees for dispensing refillable prescriptions to patients with chronic illnesses, and care management fees for enrolling new cases and performing follow-up/annual evaluations.1113 Enrolled individuals are identified as adhering to the first-stage diabetes P4P program through the following criteria: (1) 1 comprehensive claim report during the initial enrollment, (2) 5 follow-up visits, and (3) 2 annual evaluations. Individuals who adhered to the first-stage program are qualified to participate in the second-stage program. The first-stage and second-stage diabetes P4P programs involve the same important healthcare components: taking comprehensive medical histories, performing needed physical examinations/laboratory evaluations, initiating/evaluating management plans, and providing selfmanagement education. Individuals with diabetes can voluntarily decide whether they want to participate in the program or not after they receive a full description about the program by P4P-participating physicians. Neither the first-stage nor the second-stage diabetes P4P program is mandatory.11 Compared with control groups, individuals who enrolled in the diabetes P4P program showed significant improvements in the laboratory test results14 and quality of life,15 decreases in the average rate of chronic complications,15 overall medical expenditures and hospitalization costs,16 and better compliance with self-care and better satisfaction with the quality of care.13 Analyses of Taiwan’s NHI claims database showed that (1) individuals in the diabetes P4P program showed significant increases in regular follow-up visits/utilization of evidence-based services and significantly lower hospitalization costs/ total medical costs,1719 (2) although individuals in the diabetes P4P program spent more on overall healthcare expenses than the comparison group in the first year after enrollment, healthcare expenses for the comparison group were higher in the subsequent 3 years,17 (3) individuals in the diabetes P4P program or those being treated by P4Pparticipating physicians received more-comprehensive guidelinerecommended tests/examinations than those who were not in the program,8,20 and (4) individuals in the diabetes P4P program showed a lower risk of hospital admission and better medication adherence/ cost-saving than those who were not in the program.8,18 Controversial findings were also reported. Diabetes-related or overall healthcare costs for individuals in the diabetes P4P program were significantly higher than those of the controls.19,21 Systematic reviews concluded that effects of P4P programs might range from absent or negligible to strong beneficial.22,23 Inconsistent findings and a lack of long-term evaluations of P4P23,24 led us to conduct further analyses of the effects of the diabetes P4P program. A
previous longitudinal examination of Taiwan’s NHI claims database (2005e2009) used program enrollment or participation and staying in the program as a classification criterion to conduct comparisons of program effects regardless of disease duration.17 Our analyses included individuals who were all newly diagnosed with diabetes in 2001 and individuals’ longitudinal data from 2001 to 2011. Enrolled participants were classified into 2 groups: those who did and those who did not adhere to the first-stage diabetes P4P program. We aimed to identify factors related to program adherence because (1) few studies had been conducted to investigate possible factors related to program adherence, (2) program adherence might play an essential role on the estimations of program effects, and (3) adherence had a beneficial effect on quality of life for individuals with diabetes.25 Study objectives were to (1) compare the effects of the diabetes P4P program on diabetes-related/nondiabetes-related healthcare utilization/expenses between participants who did adhere to the program and those who did not, and (2) explore factors related to program adherence. Methods Design and Ethical Consideration This study was a secondary data analysis with a natural experimental design. Ethical approval was obtained from the authors’ institute. Data Source The NHI research database (2001e2011) of newly diagnosed patients with diabetes in 2001 was used for the analyses. Insured people were defined as individuals with diabetes who satisfied one of the following conditions: (1) had been hospitalized with a diabetes diagnosis or had received hypoglycemic agents during hospitalization, (2) had an outpatient visit with a diabetes diagnosis twice in a year, or (3) had 1 outpatient visit with a diabetes diagnosis and received 1 prescription for an oral hypoglycemic medication. Insured individuals who had a diabetes diagnosis in 2001 and did not use diabetes-related healthcare services in the previous 3 years were defined as individuals who were newly diagnosed with diabetes in 2001.26 The database under analyses contained 119,970 patients randomly sampled by Taiwan’s National Health Research Institutes from the total of 168,904 patients who were newly diagnosed with diabetes in 2001.26 Selection of Participants All participants included in the analyses were at least 18 years old. We did not include the patients who were enrolled in the diabetes P4P program in 2002 because Taiwan’s P4P program was implemented in November 2001, and the utilization/expenses of healthcare services at the baseline were estimated by patients’ utilization/expenses in the previous 12 months before enrollment. We accumulated 4 cohorts (n ¼ 5592) who were enrolled in the diabetes P4P program in 2003 (n ¼ 956), 2004 (n ¼ 1270), 2005 (n ¼ 1608), and 2006 (n ¼ 1460), respectively. Information about these participants’ healthcare utilization/expenses in the subsequent 5 years after their enrollment was obtained. The indices of enrolling and adherence were defined as the dates when a new case report and a second annual evaluation report were claimed, respectively. After a 3-year follow-up of the enrolled patients, 2647 (47.3%) of them did adhere to the first-stage diabetes P4P program. To minimize the differences between the characteristics of the patients who adhered to the program and those who did not, propensity score matching was adopted. The covariates included the age, gender, and
M.-J. Chi et al. / JAMDA xxx (2016) 1e7
Charlson Comorbidity Index (CCI; excluding diabetes) of the paticipants.27 We employed the Mahalanobis distance calculation method with one-to-one matches between the adherence group and nonadherence group based on the propensity score. A total of 5294 patients (adherence: 2647 vs nonadherence: 2647) were included for analyses. Measured Variables We estimated utilization/expenses of healthcare services for both groups at 6 time points (T0eT5), based on each participant’s enrollment index date at the baseline of enrollment (T0, utilization/expenses in the previous 12 months before enrollment), and at the ends of the first, second, third, fourth, and fifth years (T1-T5) after enrollment. Indicators of the utilization of healthcare services included the number/expense of diabetes-related physician visits, number/length of stays/expense of diabetes-related hospitalizations, and the annual diabetes-related healthcare expense (including outpatient, inpatient, and emergency services and other healthcare services covered by Taiwan’s NHI). The total annual healthcare expense (diabetes-related/ nondiabetes-related) was also estimated to explore the trend of changes in participants’ utilization/expenses of healthcare services. Covariates included the age, gender, diabetes-related complications, CCI, Continuity of Care Index (COCI) of the participants, time since diagnosis of diabetes, hospitalization in the previous year, and
3
location receiving healthcare services. Diabetes-related complications were defined as 7 diabetes-related complications listed in the Diabetes Complications Severity Index.28 The CCI and COCI were estimated based on information obtained from participants’ medical records in a 1-year period before their enrollment. A higher CCI score indicates that an individual experienced more comorbidities.27 The COCI that ranged from 0 to 1 was estimated based on an individual’s physician visits.29 A higher COCI score indicates that an individual received a better continuity of care and generally received healthcare services at the same location. Based on the distribution of participants’ COCI, we classified the participants into 3 groups: low (0.00e0.22), moderate (0.23e0.43), and high (0.44e1.00). The time since diagnosis of diabetes (in years) was defined as the period from the incidence year (2001) to the year of program enrollment. Statistical Analyses We analyzed the database using SAS v 9.2 (SAS Institute, Cary, NC). A P value of <.05 was considered significant. We applied t tests to test each utilization and expense of healthcare services at 6 time points between the 2 groups. A repeated-measures analysis of variance was applied to examine changes in the annual diabetes-related healthcare service expenses and total annual expenses by group. Logistic regression was used to examine factors related to program adherence.
Table 1 Individual Characteristics of Participants at the Beginning of Diabetes P4P Program Enrollment (n ¼ 5294) Variables
Adherence Group (n ¼ 2647)
Age (range, 18e99) (mean SD, years) 65 years old Yes Gender Male Diabetes-related complications Yes Cardiovascular disease Neuropathy Nephropathy Cerebrovascular disease Retinopathy Peripheral vascular disease Metabolic disorders CCI 0 1 2 COCI (median) 0.00e0.22 low 0.23e0.43 moderate 0.44e1.00 high Missing Time since diagnosis of diabetes (mean SD, years) Hospitalization in the previous year Yes Location receiving healthcare services Taipei region Central Taiwan Southern Taiwan Kaoping region Northern Taiwan Eastern Taiwan Missing Year of program enrollment 2003 2004 2005 2006
57.9 11.3
n
Nonadherence Group (n ¼ 2647) %
n
%
57.8 11.8
Total (n ¼ 5294) n
%
57.8 11.6
777
49.9
779
50.1
1556
29.4
1283
48.8
1346
51.2
2629
49.7
967 382 298 227 106 112 92 34
49.4 48.4 51.3 51.9 43.6 47.9 43.4 43.6
990 407 283 210 137 122 120 44
50.6 51.6 48.7 48.1 56.4 52.1 56.6 56.4
1957 789 581 437 243 234 212 78
37.0 14.9 11.0 8.3 4.6 4.4 4.0 1.5
49.4 51.9 49.7
2898 1107 1289 0.36 1164 2059 2060 11 3.67 1.06
54.8 20.9 24.3
1465 533 649 0.38 489 1029 1125 4 3.71 1.04
50.6 48.1 50.3 42.0 50.5 54.6 36.4
1433 574 640 0.33 675 1030 935 7 3.64 1.08
58.0 50.0 45.4 63.6
22.0 38.9 38.9 0.2
405
44.6
503
55.4
908
17.2
890 855 303 274 256 51 18
51.7 51.9 40.6 39.0 42.2 43.2 36.7
751 719 397 380 312 64 24
48.3 48.1 59.4 61.0 57.8 56.8 63.3
1641 1574 700 654 568 115 42
31.0 29.7 13.2 12.4 10.7 2.2 0.8
432 641 837 737
45.2 50.5 52.1 50.5
524 629 771 723
54.8 49.5 47.9 49.5
956 1270 1608 1460
18.0 24.0 30.4 27.6
4
M.-J. Chi et al. / JAMDA xxx (2016) 1e7
annual healthcare expenses at 6 time points (T0eT5) used by the nonadherence group were all higher than those of the adherence group (F ¼ 7.131, P < .0001).
Results Table 1 shows individual characteristic of all participants and by group. At the baseline (T0), there were no significant differences between the adherence group and nonadherence group on age, gender, and CCI.
Factors Related to Program Adherence A logistic regression showed that participants who were more likely to adhere to the program were female [odds ratio (OR) 1.16, 95% confidence interval (CI) 1.04e1.29], had higher COCI scores (moderate vs low: OR 1.37, 95% CI 1.18e1.59; high vs low: OR 1.65, 95% CI 1.42e1.92), and had a longer duration of diabetes diagnosis (OR 1.08, 95% CI 1.02e1.14). Participants who had been hospitalized in the previous year before enrollment (OR 0.79, 95% CI 0.68e0.92) were less likely to adhere to the program. Compared with participants receiving healthcare services in the Taipei region, the most intensive medical resource area in Taiwan,31 participants receiving healthcare services in the other 3 regions were less likely to adhere to the program. Logistic regression models for the 2 age groups (<65 and 65 years) were also tested. Similar findings were obtained in the age group of <65 years (Table 3).
Comparison of Healthcare Utilization/Expenses The utilization/expenses of healthcare services are shown in Table 2. At 6 time points (T0eT5), the average number and expense of physician visits for the adherence group were significantly higher than those for the nonadherence group; the average number of hospitalizations for the adherence group was significantly lower than that for the nonadherence group. At 4 time points (T0eT3), the average length of stay and expense of hospitalizations for the adherence group were significantly lower than those of the nonadherence group. The total annual healthcare expense of the adherence group was significantly lower than that of the nonadherence group at 6 time points (T0eT5). After controlling for possible influencing factors, such as age, number of diabetes-related complications, CCI, and COCI at the baseline (T0), changes in the annual diabetes-related healthcare expenses were calculated and are shown in Figure 1, A (F ¼ 8.623, P < .0001). The annual diabetes-related healthcare expenses for the 2 groups were similar before their enrollment. The expenses incurred by the nonadherence group had greatly increased and were significantly higher than those of the adherence group (T1: New Taiwan [NT]$38,277 vs NT$28,071, P < .0001; T2: NT$43,526 vs NT$29,382, P < .0001; the average exchange rate was US$1.00 z NT$ 33.43 during 2003e2006).30 The difference in the annual diabetes-related healthcare expense between groups had decreased at T3. The annual diabetes-related healthcare expenses at T4 and T5 used by the adherence group were higher than those of the nonadherence group. Figure 1, B shows changes in total annual healthcare expenses after controlling for possible influencing factors. Although the estimation lines in Figure 1, B show similar trends as those in Figure 1, A, the total
Discussion Comparison of Healthcare Utilization/Expenses Systematic reviews of the current evidence concluded that P4P have the potential to be cost-effective.23,32 Fagan et al33 analyzed claims files (2004e2007) and did not identify beneficial effects of a diabetes P4P program on healthcare utilization/costs. Curtin et al34 identified a positive return of a diabetes P4P program on investment. An examination of Taiwan’s NHI database revealed that diabetes P4P program was cost-effective and provided a return on investment of about 1.8e2.0:1.18 Analyses of Taiwan’s NHI claims database (2004e2006) showed that the diabetes P4P program was associated with a significant increase in healthcare utilization and a decrease in hospitalization
Table 2 Comparison of the Utilization/Expenses of Healthcare Services Between Groups (n ¼ 5294) (in NT$) Utilization/Expenses of Healthcare Services
Adherence Group Diabetes-related healthcare* Number of physician visits Expense of physician visits Number of hospitalizations Length of stay of hospitalizations Expense of hospitalizations Annual healthcare expense Total annual healthcare expense (diabetes-related and nondiabetes-related)* Nonadherence Group Diabetes-related healthcare* Number of physician visits Expense of physician visits Number of hospitalizations Length of stay of hospitalizations Expense of hospitalizations Annual healthcare expense Total annual healthcare expense* (diabetes-related and non-diabetes-related)
Data Collection Time Points T0
T1
T2
T3
T4
T5
12.80y 18,715y 0.15y 1.11y 5588y 24,535 42,488z
14.00y 23,162y 0.12y 0.74y 4719y 28,071y 44,755y
13.80y 24,299y 0.13y 0.79y 4869y 29,382y 45,989y
13.74y 24,948y 0.14y 1.03y 7041y 32,337z 50,218y
13.71y 24,606y 0.20y 1.88 11,886 36,845 56,377y
13.57y 24,015y 0.24z 2.12 13,971 38,491 62,790x
10.8 16,178 0.21 1.60 8650 25,191 45,845
12.63 19,889 0.30 3.04 17,914 38,277 65,318
11.62 19,170 0.33 3.70 23,743 43,526 73,926
11.21 18,367 0.29 3.07 16,793 35,671 66,141
11.13 18,902 0.27 2.46 15,619 35,075 69,416
10.87 18,856 0.29 2.80 17,518 36,935 73,245
The exchange rate of the NT$ against the US$ (2003w2006) was 1: 33.43 [Central Bank of the Republic of China (Taiwan). Foreign exchange regime. Available at: http://www. cbc.gov.tw/content.asp?CuItem¼1879]. *t tests were applied to test each utilization/expense of healthcare services at 6 time points between groups. y P < .0001. z P < .05. x P < .01.
M.-J. Chi et al. / JAMDA xxx (2016) 1e7
5
Fig. 1. (A) Changes in annual diabetes-related healthcare expenses. (B) Changes in total annual healthcare expenses (diabetes-related and nondiabetes-related).
costs. The overall healthcare costs incurred by patients in the diabetes P4P program were significantly higher than those of patients who were not enrolled.19 Cheng et al17 used Taiwan’s NHI claims data (2004e2009) to create 2 sets of samples based on patients’ enrollment status: (1) one sample set included all patients who were enrolled in the P4P program in 2005 regardless of their enrollment status after 2005 as the intervention group
(n ¼ 20,934), and (2) the other sample set included patients who were enrolled in the P4P program in 2005 and stayed in the program during the observation period as the intervention group (n ¼ 9694). We use the study results of Cheng et al17 generated from the second sample set for further discussions because these estimations might be more conservative than those generated from the first sample set.
6
M.-J. Chi et al. / JAMDA xxx (2016) 1e7
Table 3 Factors Related to Program Adherence (n ¼ 5241) Variables
Program Adherence (ref: no) Total (n ¼ 5241)
Age <65 years (n ¼ 3695)
Age 65 years (n ¼ 1656)
OR
OR
OR
95% CI
-
-
95% CI
95% CI
Gender (ref: male) Female 1.16* 1.04e1.29 1.17* 1.03e1.34 COCI (ref: 0.00e0.22) 0.23e0.43 1.37y 1.18e1.59 1.41y 1.18e1.68 moderate 0.44e1.00 high 1.65y 1.42e1.92 1.60y 1.34e1.91 1.08z 1.02e1.14 1.10z 1.03e1.17 Time since diagnosis of diabetes (years) Hospitalization in the previous year (ref: no) Yes 0.79z 0.68e0.92 0.76z 0.63e0.92 Location receiving healthcare services (ref: Taipei region) Central Taiwan 1.07 0.93e1.24 1.15 0.97e1.37 Southern Taiwan 0.69y 0.58e0.83 0.80* 0.65e0.99 The Kaoping 0.65y 0.54e0.78 0.64y 0.52e0.80 region Northern Taiwan 0.73z 0.60e0.89 0.72z 0.57e0.91 Eastern Taiwan 0.70 0.48e1.03 0.72 0.46e1.13
1.31 1.83 -
0.99e1.72 y
1.38e2.43 -
-
-
0.92 0.49y 0.69*
0.71e1.19 0.34e0.68 0.49e0.98
0.78 0.64
0.55e1.10 0.31e1.33
Logistic regression with conditional forward method was used to explore the factors related to program adherence. The independent variables included gender, age 65 years, number of diabetes-related complications, CCI, COCI, time since diagnosis of diabetes, hospitalization in the previous 1 year, and location receiving healthcare services. *P < .05. y P < .0001. z P < .01
Our study findings were similar to the findings of Lee et al19 and Cheng et al.17 Healthcare utilization/expenses for patients in the P4P program increased in the initial period after enrollment. Long-term, beneficial effects of the P4P program might have been present if patients had adhered to the program at certain levels. The reported long-term, beneficial effects might imply that avoidable healthcare expenses can be diminished because patients’ conditions and possible problems are regularly monitored, detected, and managed through comprehensive follow-up/evaluations. However, further evidence related to the long-term effects of diabetes P4P programs on healthcare expenses and potential savings is still needed. The expenses of hospitalizations (T1eT3) for the adherence group were significantly less than those for the nonadherence group. This study finding implies that the total healthcare expense saving incurred in the adherence group might be related to the cost savings on the expense of hospitalizations. Further exploring the potential areas of significant cost savings (eg, the expense of hospitalizations) resulted from the P4P program is recommended. Factors Related to Program Adherence Few studies have been conducted to investigate factors related to P4P program adherence. Thus, we describe factors related to healthcare utilization/costs, and adherence to practice guidelines/dispensed medication among patients with diabetes for further discussions. Significantly lower healthcare utilization was observed in men compared with women in the numbers of physician/practitioner/ medical specialist visits and laboratory tests.35 Better continuity of care was significantly associated with fewer hospital/emergency visit36 and hospitalizations,8,37 and decreased healthcare costs.8,37 Renard et al38 found that (1) men were less likely to adhere to recommended guidelines, and (2) patients who had been admitted to the hospital in the past or current year better adhered to guidelines than did those who had not been admitted. Satman et al39 found that
patients with longer durations of diabetes had better guideline adherence. Analyses of Taiwan’s NHI claims data showed that patients in the diabetes P4P program or those treated by P4P-participating physicians were more likely to adhere to practice guidelines than those who were treated by non-P4P-participating physicians.20 Patients with longer durations of diabetes and higher continuity of care scores were more likely to have good medication adherence.8,40 Gender, continuity of care, and time since diagnosis of diabetes were identified as factors related to healthcare utilization/costs, and adherence to practice guidelines/dispensed medications.8,3540 In this study, relationships between program adherence and these factors were found. Findings presented in this study and previous studies suggest that healthcare professionals should pay special attention to patients who are male and newly diagnosed with diabetes, and who have low continuity of care scores. Further studies including interventions or educational strategies which could improve patients’ adherence and continuity of care level are suggested to achieve optimal effects, disease control and clinical outcomes. A previous study found that patients who had been admitted to the hospital possessed better practice guideline adherence.38 Gibson et al41 found that higher levels of medication adherence were associated with more physician visits and fewer inpatient admissions/ emergency visits. The study results of Gibson et al41 might explain the relationships between program adherence and diabetes-related hospitalization identified in this study. Nuti et al42 reported that no-shows to a primary care appointment were associated with increased risks of hospitalization among patients with diabetes who had recently been hospitalized. Further studies including management strategies that could improve patients’ adherence are suggested. Feasible strategies include (1) providing needed education/counseling, (2) increasing the convenience of provided care, (3) improving patients’ involvement, and (4) using reminders, reinforcement, or rewards.43 One study conducted in a rural area of Taiwan found that rural residents with diabetes tended to exhibit inadequate health-related behaviors.44 In this study, the mean COCI score for participants receiving healthcare services in the Taipei region (0.43) was significantly higher than those of participants receiving services in southern Taiwan (0.40), the Kaoping region (0.39), and eastern Taiwan (0.36). Better program adherence of participants receiving healthcare services in the Taipei region might be explained by participants living in an urban area43 and those with higher continuity of care scores40 being more likely to perform healthcare behaviors. There are several limitations of this study. First, we were unable to control for other potentially important variables (eg, ethnicity, education, cognitive status, and number/type of medication usage), although we did control for available confounders when we conducted the analyses. Second, this study could not analyze some outcome measures of diabetes care (eg, the hemoglobin A1C) because of a lack of information. Third, we used the administrative definition of adherence to the first-stage P4P program. It is possible that some cases may have been misclassified. Achieving the first-stage P4P program goal for patients who were not meeting glycemic goals might not be sufficient to be defined as delivering quality care.45 Fourth, the cohort is representative of the population of Taiwan. The ability to generalize our study results might be limited because of different settings and different designs of P4P programs. Fifth, improved outcomes with a decrease in healthcare spending cannot completely represent a high quality of care. Economic evaluations with controlled intervention designs are suggested to identify more precise estimations. Conclusions Our study results provide further evidence of the long-term, beneficial effects of the diabetes P4P program, although healthcare utilization/expenses cannot completely represent the delivery of
M.-J. Chi et al. / JAMDA xxx (2016) 1e7
quality care. Study results also showed that the total healthcare expense saving incurred in the adherence group might be related to the cost saving on the expense of hospitalizations. Further exploration of the long-term effects of diabetes P4P programs on healthcare expenses and the potential areas of significant healthcare cost savings is recommended. Gender, continuity of care, time since diagnosis of diabetes, hospitalization in the previous year, and location receiving healthcare services were identified as factors related to program adherence. Further studies including intervention, education, or management strategies, which can improve patients’ adherence and continuity of care levels, are suggested to achieve optimal effects, disease control, and clinical outcomes. Acknowledgments We greatly appreciate the individuals who provided information for our analyses and the financial support of College of Nursing, Taipei Medical University. References 1. Danaei G, Finucane MM, Lu Y, et al. National, regional, and global trends in fasting plasma glucose and diabetes prevalence since 1980: Systematic analysis of health examination surveys and epidemiological studies with 370 countryyears and 2.7 million participants. Lancet 2011;378:31e40. 2. Lozano R, Naghavi M, Foreman K, et al. Global and regional mortality from 235 causes of death for 20 age groups in 1990 and 2010: A systematic analysis for the Global Burden of Disease Study 2010. Lancet 2012;380:2095e2128. 3. Ministry of Health and Welfare. 2015 Causes of death statistics. Available at: http://www.mohw.gov.tw/cht/DOS/Statistic.aspx?f_list_no¼312&fod_list_no¼ 5012. Accessed October 25, 2015. 4. National Health Insurance Administration. National Health Insurance in Taiwan 2014-2015 annual report. Available at: http://www.nhi.gov.tw/ webdata/webdata.aspx?menu¼17&menu_id¼659&webdata_id¼2891&WD_ID¼ 897. Accessed October 25, 2015. 5. Jiang YD, Chang CH, Tai TY, et al. Incidence and prevalence rates of diabetes mellitus in Taiwan: Analysis of the 2000e2009 Nationwide Health Insurance database. J Formos Med Assoc 2012;111:599e604. 6. International Diabetes Federation. IDF diabetes atlas, 2013 (6th ed.). Available at: http://www.idf.org/diabetesatlas. Accessed October 25, 2015. 7. Chang K. Comorbidities, quality of life and patients’ willingness to pay for a cure for type two diabetes in Taiwan. Public Health 2010;124:284e294. 8. Cheng JS, Tsai WC, Lin CL, et al. Trend and factors associated with healthcare use and costs in type two diabetes mellitus: A decade experience of a universal health insurance program. Med Care 2015;53:116e124. 9. Chiou ST, Lin HD, Yu NC, et al. An initial assessment of the feasibility and effectiveness of implementing diabetes shared care system in Taiwan: Some experiences from I-Lan County. Diabetes Res Clin Pract 2001;54:S67eS73. 10. Rosenthal MB, Fernandopulle R, Song HR, Landon B. Paying for quality: Providers’ incentives for quality improvement. Health Aff 2004;23:127e141. 11. National Health Insurance Administration. 2011 National Health Insurance Pay-for-Performance program for diabetes (in Chinese). Available at: http:// www.nhi.gov.tw/webdata/webdata.aspx?menu¼20&menu_id¼712&webdata_ id¼3862&WD_ID¼836. Accessed October 25, 2015. 12. Chang RE, Lin SP, Aron DC. A pay-for-performance program in Taiwan improved care for some diabetes patients, but doctors may have excluded sicker ones. Health Aff 2012;31:93e102. 13. Chen PC, Lee YC, Kuo RN. Differences in patient reports on the quality of care in a diabetes pay-for-performance program between 1 year enrolled and newly enrolled patients. Int J Qual Health Care 2012;24:189e196. 14. Chen TW, Wang JY. Effectiveness of the diabetes shared care network for middle-aged and elderly patients in a rural area of Taiwan. Taiwan J Public Health 2009;28:334e343. 15. Lee CP, Chu CI, Wu DA. Comparison of the diabetes regular care model and the “Diabetes shared care disease management” model in eastern Taiwan. Tzu Chi Med J 2007;19:66e73. 16. Hsu HC, Weng HC, Lin YT, et al. A study of economic clinical and satisfaction evaluation of effectiveness of diabetes disease management program: A pilot study of a community hospital. J Health Manag (Taiwan) 2004;5:222e242. 17. Cheng SH, Lee TT, Chen CC. A longitudinal examination of a pay-forperformance program for diabetes care: Evidence from a natural experiment. Med Care 2012;50:109e116.
7
18. Hsieh HM, Tsai SL, Shin SJ, et al. Cost-effectiveness of diabetes pay-forperformance incentive designs. Med Care 2015;53:106e115. 19. Lee TT, Cheng SH, Chen CC, Lai MS. A pay-for-performance program for diabetes care in Taiwan: A preliminary assessment. Am J Manag Care 2010;16: 65e69. 20. Lai CL, Hou YH. The association of clinical guideline adherence and pay-forperformance among patients with diabetes. J Chinese Med Assoc 2013;76: 102e107. 21. Tan EC, Pwu RF, Chen DR, Yang MC. Is a diabetes pay-for-performance program cost-effective under the National Health Insurance in Taiwan? Qual Life Res 2014;23:687e696. 22. Huang J, Yin S, Lin Y, et al. Impact of pay-for-performance on management of diabetes: A systematic review. J Evid Based Med 2013;6:173e184. 23. Van Herck P, De Smedt D, Annemans L, et al. Systematic review: Effects, design choices, and context of pay-for-performance in health care. BMC Health Serv Res 2010;10:247. 24. Scott A, Sivey P, Ait Ouakrim D, et al. The effect of financial incentives on the quality of health care provided by primary care physicians. Cochrane Database Syst Rev 2011:CD008451. 25. Chew BH. Medication adherence on quality of life among adults with type 2 diabetes mellitus: An exploratory analysis on the EDDMQoL study. Qual Life Res 2015;24:2723e2731. 26. National Health Research Institutes (NHRI). 2014 National Health Insurance Research Database: Longitudinal cohort of diabetes patients. Available at: http://nhird.nhri.org.tw/date_lhdb.htm. Accessed July 15, 2014. 27. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: Development and validation. J Chronic Dis 1987;40:373e383. 28. Young BA, Lin E, Von Korff M, et al. Diabetes complications severity index and risk of mortality, hospitalization, and healthcare utilization. Am J Manag Care 2008;14:15e23. 29. Bice TW, Boxerman SB. A quantitative measure of continuity of care. Med Care 1977;15:347e349. 30. Central Bank of the Republic of China (Taiwan). 2015 Foreign exchange regime. Available at: http://www.cbc.gov.tw/content.asp?CuItem¼1879. Accessed September 20, 2015. 31. Ministry of Health and Welfare. Number of hospitals, clinics and beds by category and locale, 2008-2013. Available at: http://www.mohw.gov.tw/EN/ Ministry/Statistic.aspx?f_list_no¼474&fod_list_no¼5173. Accessed September 20, 2015. 32. Eijkenaar F, Emmert M, Scheppach M, Schöffski O. Effects of pay for performance in health care: A systematic review of systematic reviews. Health Policy 2013;110:115e130. 33. Fagan PJ, Schuster AB, Boyd C, et al. Chronic care improvement in primary care: Evaluation of an integrated pay-for-performance and practice-based care coordination program among elderly patients with diabetes. Health Serv Res J 2010;45:1763e1782. 34. Curtin K, Beckman H, Pankow G, et al. Return on investment in pay for performance: A diabetes case study. J Healthc Manag 2006;51:365e374. discussion 375e376. 35. Krämer HU, Rüter G, Schöttker B, et al. Gender differences in healthcare utilization of patients with diabetes. Am J Manag Care 2012;18:362e369. 36. Chen CC, Cheng SH. Better continuity of care reduces costs for diabetic patients. Am J Manag Care 2011;17:420e427. 37. Hong JS, Kang HC. Continuity of ambulatory care and health outcomes in adult patients with type 2 diabetes in Korea. Health Policy 2013;109: 158e165. 38. Renard LM, Bocquet V, Vidal-Trecan G, et al. Adherence to international followup guidelines in type two diabetes: A longitudinal cohort study in Luxembourg. PLoS One 2013;8:e80162. 39. Satman I, Imamoglu S, Yilmaz C, ADMIRE Study Group. A patient-based study on the adherence of physicians to guidelines for the management of type two diabetes in Turkey. Diabetes Res Clin Pract 2012;98:75e82. 40. Chang HY, Chiou CJ, Lin MC, et al. A population study of the self-care behaviors and their associated factors of diabetes in Taiwan: Results from the 2001 National Health Interview Survey in Taiwan. Prev Med 2005;40: 344e348. 41. Gibson TB, Song X, Alemayehu B, et al. Cost sharing, adherence, and health outcomes in patients with diabetes. Am J Manag Care 2010;16:589e600. 42. Nuti LA, Lawley M, Turkcan A, et al. No-shows to primary care appointments: Subsequent acute care utilization among diabetic patients. BMC Health Serv Res 2012;12:304. 43. Lerman I. Adherence to treatment: The key for avoiding long-term complications of diabetes. Arch Med Res 2005;36:300e306. 44. Chen MY, Huang WC, Peng YS, et al. Health status and health-related behaviors among type 2 diabetes community residents. J Nurs Res 2011;19:35e43. 45. American Diabetes Association. Standards of medical care in diabetes-2014. Diabetes Care 2014;37:S14eS80.
JAMDA journal homepage: www.jamda.com
Original Study
Effects and Factors Related to Adherence to A Diabetes Pay-for-Performance Program: Analyses of a National Health Insurance Claims Database Mei-Ju Chi PhD a, b, Kuei-Ru Chou PhD c, d, e, Dee Pei MD, PhD f, g, Jawl-Shan Hwang MD h, i, Laurie Quinn PhD j, Min-Huey Chung PhD c, Yuan-Mei Liao PhD c, * a
School of Gerontology Health Management, College of Nursing, Taipei Medical University, Taipei City, Taiwan Master Program in Long-term Care, College of Nursing, Taipei Medical University, Taipei City, Taiwan School of Nursing, College of Nursing, Taipei Medical University, Taipei City, Taiwan d Psychiatric Research Center, Taipei Medical University Hospital, Taipei City, Taiwan e Department of Nursing, Taipei Medical University-Shuang Ho Hospital, New Taipei City, Taiwan f College of Medicine, Fu Jen Catholic University, New Taipei City, Taiwan g Department of Endocrinology, Cardinal Tien Hospital, New Taipei City, Taiwan h Division of Endocrinology and Metabolism, Department of Internal Medicine, Chang Gung Memorial Hospital-Linkou, Taoyuan City, Taiwan i College of Medicine, Chang Gung University, Taoyuan City, Taiwan j Department of Biobehavioral Health Science, College of Nursing, University of Illinois at Chicago, Chicago, IL b c
a b s t r a c t Keywords: adherence diabetes healthcare expense healthcare utilization pay-for-performance
Objectives: To compare the effects of a diabetes pay-for-performance (P4P) program on diabetes-related/ nondiabetes-related healthcare utilization/expenses between participants who adhered to the program and those who did not, and explore factors related to program adherence. Design: A secondary data analysis with a natural experimental design. Setting: Taiwan’s National Health Insurance claims database (2001e2011) of newly diagnosed patients with diabetes in 2001 was used for the analyses. Participants: The database under analyses contained 119,970 patients who were newly diagnosed with diabetes in 2001. Longitudinal data from 2001 to 2011 were obtained. A sample of 5592 patients who were enrolled in the diabetes P4P program during 2003e2006 was identified. After a 3-year follow-up of the enrolled patients, 2647 (47.3%) of them adhered to the program. To minimize the differences between the characteristics of the patients who adhered to the program and those who did not, propensity score matching was adopted. A total of 5294 patients (adherence: 2647 vs nonadherence: 2647) were included for analyses. Measurements: We estimated utilization/expenses of healthcare services for both groups at 6 time points and applied t tests to test each utilization and expense of healthcare services between the 2 groups. A repeated-measures analysis of variance was applied to examine changes in the annual diabetes-related healthcare service expenses and total annual expenses by group. Logistic regression models were used to examine factors related to program adherence. Covariates included participant age, gender, diabetesrelated complications, Charlson Comorbidity Index, Continuity of Care Index, time since diagnosis of diabetes, hospitalization in the previous year, and location receiving healthcare services. Results: Total annual healthcare expenses spent by the adherence group were significantly lower than those of the nonadherence group. Gender, continuity of care, time since diagnosis of diabetes, hospitalizations in the previous year, and location receiving healthcare services were factors related to program adherence. Conclusions: Long-term, beneficial effects of the diabetes P4P program might have been present if patients had adhered to the program. Interventions and strategies which could improve program adherence and continuity of care are suggested to achieve optimal disease control and clinical outcomes. Ó 2016 AMDA e The Society for Post-Acute and Long-Term Care Medicine.
This work was supported by a grant from the College of Nursing, Taipei Medical University. The authors declare no conflicts of interest.
http://dx.doi.org/10.1016/j.jamda.2016.02.033 1525-8610/Ó 2016 AMDA e The Society for Post-Acute and Long-Term Care Medicine.
* Address correspondence to Yuan-Mei Liao, PhD, School of Nursing, College of Nursing, Taipei Medical University, 250 Wu-Xing St, Xinyi District, Taipei City, Taiwan 110. E-mail address: [email protected] (Y.-M. Liao).
2
M.-J. Chi et al. / JAMDA xxx (2016) 1e7
Diabetes affects more than 300 million individuals globally.1 Diabetes is one of the top 10 causes of death worldwide2; it is Taiwan’s fourth or fifth leading cause of death.3 The population coverage rate of Taiwan’s National Health Insurance (NHI) has reached 99.9%.4 Analyses of Taiwan’s NHI claims database revealed that the prevalence of diabetes increased 35% from 2000 to 2009, and the total population with diabetes increased more than 70%.5 Health spending on diabetes accounted for 10.8% of the worldwide total health expenditure.6 In Taiwan, the total healthcare cost for individuals with diabetes was approximately 2.8 times of the cost spent by the age-matched and sex-matched individuals without diabetes, and the total expenditure of diabetes to society is about US$2.96 billion.7,8 In Taiwan, a diabetes shared-care program was initiated in 1996; it is a comprehensive, integrated approach to provide healthcare and reimbursement based on a disease’s natural course.9 Interest in payfor-performance (P4P) as a strategy to stimulate the delivery of quality care is increasing; P4P provides financial rewards to healthcare providers for quality improvements.10 To enhance the operation of Taiwan’s diabetes shared-care program, a diabetes P4P program was implemented in 2001. In Taiwan’s diabetes P4P program, financial incentives include additional physician fees for providing comprehensive care, fees for dispensing refillable prescriptions to patients with chronic illnesses, and care management fees for enrolling new cases and performing follow-up/annual evaluations.1113 Enrolled individuals are identified as adhering to the first-stage diabetes P4P program through the following criteria: (1) 1 comprehensive claim report during the initial enrollment, (2) 5 follow-up visits, and (3) 2 annual evaluations. Individuals who adhered to the first-stage program are qualified to participate in the second-stage program. The first-stage and second-stage diabetes P4P programs involve the same important healthcare components: taking comprehensive medical histories, performing needed physical examinations/laboratory evaluations, initiating/evaluating management plans, and providing selfmanagement education. Individuals with diabetes can voluntarily decide whether they want to participate in the program or not after they receive a full description about the program by P4P-participating physicians. Neither the first-stage nor the second-stage diabetes P4P program is mandatory.11 Compared with control groups, individuals who enrolled in the diabetes P4P program showed significant improvements in the laboratory test results14 and quality of life,15 decreases in the average rate of chronic complications,15 overall medical expenditures and hospitalization costs,16 and better compliance with self-care and better satisfaction with the quality of care.13 Analyses of Taiwan’s NHI claims database showed that (1) individuals in the diabetes P4P program showed significant increases in regular follow-up visits/utilization of evidence-based services and significantly lower hospitalization costs/ total medical costs,1719 (2) although individuals in the diabetes P4P program spent more on overall healthcare expenses than the comparison group in the first year after enrollment, healthcare expenses for the comparison group were higher in the subsequent 3 years,17 (3) individuals in the diabetes P4P program or those being treated by P4Pparticipating physicians received more-comprehensive guidelinerecommended tests/examinations than those who were not in the program,8,20 and (4) individuals in the diabetes P4P program showed a lower risk of hospital admission and better medication adherence/ cost-saving than those who were not in the program.8,18 Controversial findings were also reported. Diabetes-related or overall healthcare costs for individuals in the diabetes P4P program were significantly higher than those of the controls.19,21 Systematic reviews concluded that effects of P4P programs might range from absent or negligible to strong beneficial.22,23 Inconsistent findings and a lack of long-term evaluations of P4P23,24 led us to conduct further analyses of the effects of the diabetes P4P program. A
previous longitudinal examination of Taiwan’s NHI claims database (2005e2009) used program enrollment or participation and staying in the program as a classification criterion to conduct comparisons of program effects regardless of disease duration.17 Our analyses included individuals who were all newly diagnosed with diabetes in 2001 and individuals’ longitudinal data from 2001 to 2011. Enrolled participants were classified into 2 groups: those who did and those who did not adhere to the first-stage diabetes P4P program. We aimed to identify factors related to program adherence because (1) few studies had been conducted to investigate possible factors related to program adherence, (2) program adherence might play an essential role on the estimations of program effects, and (3) adherence had a beneficial effect on quality of life for individuals with diabetes.25 Study objectives were to (1) compare the effects of the diabetes P4P program on diabetes-related/nondiabetes-related healthcare utilization/expenses between participants who did adhere to the program and those who did not, and (2) explore factors related to program adherence. Methods Design and Ethical Consideration This study was a secondary data analysis with a natural experimental design. Ethical approval was obtained from the authors’ institute. Data Source The NHI research database (2001e2011) of newly diagnosed patients with diabetes in 2001 was used for the analyses. Insured people were defined as individuals with diabetes who satisfied one of the following conditions: (1) had been hospitalized with a diabetes diagnosis or had received hypoglycemic agents during hospitalization, (2) had an outpatient visit with a diabetes diagnosis twice in a year, or (3) had 1 outpatient visit with a diabetes diagnosis and received 1 prescription for an oral hypoglycemic medication. Insured individuals who had a diabetes diagnosis in 2001 and did not use diabetes-related healthcare services in the previous 3 years were defined as individuals who were newly diagnosed with diabetes in 2001.26 The database under analyses contained 119,970 patients randomly sampled by Taiwan’s National Health Research Institutes from the total of 168,904 patients who were newly diagnosed with diabetes in 2001.26 Selection of Participants All participants included in the analyses were at least 18 years old. We did not include the patients who were enrolled in the diabetes P4P program in 2002 because Taiwan’s P4P program was implemented in November 2001, and the utilization/expenses of healthcare services at the baseline were estimated by patients’ utilization/expenses in the previous 12 months before enrollment. We accumulated 4 cohorts (n ¼ 5592) who were enrolled in the diabetes P4P program in 2003 (n ¼ 956), 2004 (n ¼ 1270), 2005 (n ¼ 1608), and 2006 (n ¼ 1460), respectively. Information about these participants’ healthcare utilization/expenses in the subsequent 5 years after their enrollment was obtained. The indices of enrolling and adherence were defined as the dates when a new case report and a second annual evaluation report were claimed, respectively. After a 3-year follow-up of the enrolled patients, 2647 (47.3%) of them did adhere to the first-stage diabetes P4P program. To minimize the differences between the characteristics of the patients who adhered to the program and those who did not, propensity score matching was adopted. The covariates included the age, gender, and
M.-J. Chi et al. / JAMDA xxx (2016) 1e7
Charlson Comorbidity Index (CCI; excluding diabetes) of the paticipants.27 We employed the Mahalanobis distance calculation method with one-to-one matches between the adherence group and nonadherence group based on the propensity score. A total of 5294 patients (adherence: 2647 vs nonadherence: 2647) were included for analyses. Measured Variables We estimated utilization/expenses of healthcare services for both groups at 6 time points (T0eT5), based on each participant’s enrollment index date at the baseline of enrollment (T0, utilization/expenses in the previous 12 months before enrollment), and at the ends of the first, second, third, fourth, and fifth years (T1-T5) after enrollment. Indicators of the utilization of healthcare services included the number/expense of diabetes-related physician visits, number/length of stays/expense of diabetes-related hospitalizations, and the annual diabetes-related healthcare expense (including outpatient, inpatient, and emergency services and other healthcare services covered by Taiwan’s NHI). The total annual healthcare expense (diabetes-related/ nondiabetes-related) was also estimated to explore the trend of changes in participants’ utilization/expenses of healthcare services. Covariates included the age, gender, diabetes-related complications, CCI, Continuity of Care Index (COCI) of the participants, time since diagnosis of diabetes, hospitalization in the previous year, and
3
location receiving healthcare services. Diabetes-related complications were defined as 7 diabetes-related complications listed in the Diabetes Complications Severity Index.28 The CCI and COCI were estimated based on information obtained from participants’ medical records in a 1-year period before their enrollment. A higher CCI score indicates that an individual experienced more comorbidities.27 The COCI that ranged from 0 to 1 was estimated based on an individual’s physician visits.29 A higher COCI score indicates that an individual received a better continuity of care and generally received healthcare services at the same location. Based on the distribution of participants’ COCI, we classified the participants into 3 groups: low (0.00e0.22), moderate (0.23e0.43), and high (0.44e1.00). The time since diagnosis of diabetes (in years) was defined as the period from the incidence year (2001) to the year of program enrollment. Statistical Analyses We analyzed the database using SAS v 9.2 (SAS Institute, Cary, NC). A P value of <.05 was considered significant. We applied t tests to test each utilization and expense of healthcare services at 6 time points between the 2 groups. A repeated-measures analysis of variance was applied to examine changes in the annual diabetes-related healthcare service expenses and total annual expenses by group. Logistic regression was used to examine factors related to program adherence.
Table 1 Individual Characteristics of Participants at the Beginning of Diabetes P4P Program Enrollment (n ¼ 5294) Variables
Adherence Group (n ¼ 2647)
Age (range, 18e99) (mean SD, years) 65 years old Yes Gender Male Diabetes-related complications Yes Cardiovascular disease Neuropathy Nephropathy Cerebrovascular disease Retinopathy Peripheral vascular disease Metabolic disorders CCI 0 1 2 COCI (median) 0.00e0.22 low 0.23e0.43 moderate 0.44e1.00 high Missing Time since diagnosis of diabetes (mean SD, years) Hospitalization in the previous year Yes Location receiving healthcare services Taipei region Central Taiwan Southern Taiwan Kaoping region Northern Taiwan Eastern Taiwan Missing Year of program enrollment 2003 2004 2005 2006
57.9 11.3
n
Nonadherence Group (n ¼ 2647) %
n
%
57.8 11.8
Total (n ¼ 5294) n
%
57.8 11.6
777
49.9
779
50.1
1556
29.4
1283
48.8
1346
51.2
2629
49.7
967 382 298 227 106 112 92 34
49.4 48.4 51.3 51.9 43.6 47.9 43.4 43.6
990 407 283 210 137 122 120 44
50.6 51.6 48.7 48.1 56.4 52.1 56.6 56.4
1957 789 581 437 243 234 212 78
37.0 14.9 11.0 8.3 4.6 4.4 4.0 1.5
49.4 51.9 49.7
2898 1107 1289 0.36 1164 2059 2060 11 3.67 1.06
54.8 20.9 24.3
1465 533 649 0.38 489 1029 1125 4 3.71 1.04
50.6 48.1 50.3 42.0 50.5 54.6 36.4
1433 574 640 0.33 675 1030 935 7 3.64 1.08
58.0 50.0 45.4 63.6
22.0 38.9 38.9 0.2
405
44.6
503
55.4
908
17.2
890 855 303 274 256 51 18
51.7 51.9 40.6 39.0 42.2 43.2 36.7
751 719 397 380 312 64 24
48.3 48.1 59.4 61.0 57.8 56.8 63.3
1641 1574 700 654 568 115 42
31.0 29.7 13.2 12.4 10.7 2.2 0.8
432 641 837 737
45.2 50.5 52.1 50.5
524 629 771 723
54.8 49.5 47.9 49.5
956 1270 1608 1460
18.0 24.0 30.4 27.6
4
M.-J. Chi et al. / JAMDA xxx (2016) 1e7
annual healthcare expenses at 6 time points (T0eT5) used by the nonadherence group were all higher than those of the adherence group (F ¼ 7.131, P < .0001).
Results Table 1 shows individual characteristic of all participants and by group. At the baseline (T0), there were no significant differences between the adherence group and nonadherence group on age, gender, and CCI.
Factors Related to Program Adherence A logistic regression showed that participants who were more likely to adhere to the program were female [odds ratio (OR) 1.16, 95% confidence interval (CI) 1.04e1.29], had higher COCI scores (moderate vs low: OR 1.37, 95% CI 1.18e1.59; high vs low: OR 1.65, 95% CI 1.42e1.92), and had a longer duration of diabetes diagnosis (OR 1.08, 95% CI 1.02e1.14). Participants who had been hospitalized in the previous year before enrollment (OR 0.79, 95% CI 0.68e0.92) were less likely to adhere to the program. Compared with participants receiving healthcare services in the Taipei region, the most intensive medical resource area in Taiwan,31 participants receiving healthcare services in the other 3 regions were less likely to adhere to the program. Logistic regression models for the 2 age groups (<65 and 65 years) were also tested. Similar findings were obtained in the age group of <65 years (Table 3).
Comparison of Healthcare Utilization/Expenses The utilization/expenses of healthcare services are shown in Table 2. At 6 time points (T0eT5), the average number and expense of physician visits for the adherence group were significantly higher than those for the nonadherence group; the average number of hospitalizations for the adherence group was significantly lower than that for the nonadherence group. At 4 time points (T0eT3), the average length of stay and expense of hospitalizations for the adherence group were significantly lower than those of the nonadherence group. The total annual healthcare expense of the adherence group was significantly lower than that of the nonadherence group at 6 time points (T0eT5). After controlling for possible influencing factors, such as age, number of diabetes-related complications, CCI, and COCI at the baseline (T0), changes in the annual diabetes-related healthcare expenses were calculated and are shown in Figure 1, A (F ¼ 8.623, P < .0001). The annual diabetes-related healthcare expenses for the 2 groups were similar before their enrollment. The expenses incurred by the nonadherence group had greatly increased and were significantly higher than those of the adherence group (T1: New Taiwan [NT]$38,277 vs NT$28,071, P < .0001; T2: NT$43,526 vs NT$29,382, P < .0001; the average exchange rate was US$1.00 z NT$ 33.43 during 2003e2006).30 The difference in the annual diabetes-related healthcare expense between groups had decreased at T3. The annual diabetes-related healthcare expenses at T4 and T5 used by the adherence group were higher than those of the nonadherence group. Figure 1, B shows changes in total annual healthcare expenses after controlling for possible influencing factors. Although the estimation lines in Figure 1, B show similar trends as those in Figure 1, A, the total
Discussion Comparison of Healthcare Utilization/Expenses Systematic reviews of the current evidence concluded that P4P have the potential to be cost-effective.23,32 Fagan et al33 analyzed claims files (2004e2007) and did not identify beneficial effects of a diabetes P4P program on healthcare utilization/costs. Curtin et al34 identified a positive return of a diabetes P4P program on investment. An examination of Taiwan’s NHI database revealed that diabetes P4P program was cost-effective and provided a return on investment of about 1.8e2.0:1.18 Analyses of Taiwan’s NHI claims database (2004e2006) showed that the diabetes P4P program was associated with a significant increase in healthcare utilization and a decrease in hospitalization
Table 2 Comparison of the Utilization/Expenses of Healthcare Services Between Groups (n ¼ 5294) (in NT$) Utilization/Expenses of Healthcare Services
Adherence Group Diabetes-related healthcare* Number of physician visits Expense of physician visits Number of hospitalizations Length of stay of hospitalizations Expense of hospitalizations Annual healthcare expense Total annual healthcare expense (diabetes-related and nondiabetes-related)* Nonadherence Group Diabetes-related healthcare* Number of physician visits Expense of physician visits Number of hospitalizations Length of stay of hospitalizations Expense of hospitalizations Annual healthcare expense Total annual healthcare expense* (diabetes-related and non-diabetes-related)
Data Collection Time Points T0
T1
T2
T3
T4
T5
12.80y 18,715y 0.15y 1.11y 5588y 24,535 42,488z
14.00y 23,162y 0.12y 0.74y 4719y 28,071y 44,755y
13.80y 24,299y 0.13y 0.79y 4869y 29,382y 45,989y
13.74y 24,948y 0.14y 1.03y 7041y 32,337z 50,218y
13.71y 24,606y 0.20y 1.88 11,886 36,845 56,377y
13.57y 24,015y 0.24z 2.12 13,971 38,491 62,790x
10.8 16,178 0.21 1.60 8650 25,191 45,845
12.63 19,889 0.30 3.04 17,914 38,277 65,318
11.62 19,170 0.33 3.70 23,743 43,526 73,926
11.21 18,367 0.29 3.07 16,793 35,671 66,141
11.13 18,902 0.27 2.46 15,619 35,075 69,416
10.87 18,856 0.29 2.80 17,518 36,935 73,245
The exchange rate of the NT$ against the US$ (2003w2006) was 1: 33.43 [Central Bank of the Republic of China (Taiwan). Foreign exchange regime. Available at: http://www. cbc.gov.tw/content.asp?CuItem¼1879]. *t tests were applied to test each utilization/expense of healthcare services at 6 time points between groups. y P < .0001. z P < .05. x P < .01.
M.-J. Chi et al. / JAMDA xxx (2016) 1e7
5
Fig. 1. (A) Changes in annual diabetes-related healthcare expenses. (B) Changes in total annual healthcare expenses (diabetes-related and nondiabetes-related).
costs. The overall healthcare costs incurred by patients in the diabetes P4P program were significantly higher than those of patients who were not enrolled.19 Cheng et al17 used Taiwan’s NHI claims data (2004e2009) to create 2 sets of samples based on patients’ enrollment status: (1) one sample set included all patients who were enrolled in the P4P program in 2005 regardless of their enrollment status after 2005 as the intervention group
(n ¼ 20,934), and (2) the other sample set included patients who were enrolled in the P4P program in 2005 and stayed in the program during the observation period as the intervention group (n ¼ 9694). We use the study results of Cheng et al17 generated from the second sample set for further discussions because these estimations might be more conservative than those generated from the first sample set.
6
M.-J. Chi et al. / JAMDA xxx (2016) 1e7
Table 3 Factors Related to Program Adherence (n ¼ 5241) Variables
Program Adherence (ref: no) Total (n ¼ 5241)
Age <65 years (n ¼ 3695)
Age 65 years (n ¼ 1656)
OR
OR
OR
95% CI
-
-
95% CI
95% CI
Gender (ref: male) Female 1.16* 1.04e1.29 1.17* 1.03e1.34 COCI (ref: 0.00e0.22) 0.23e0.43 1.37y 1.18e1.59 1.41y 1.18e1.68 moderate 0.44e1.00 high 1.65y 1.42e1.92 1.60y 1.34e1.91 1.08z 1.02e1.14 1.10z 1.03e1.17 Time since diagnosis of diabetes (years) Hospitalization in the previous year (ref: no) Yes 0.79z 0.68e0.92 0.76z 0.63e0.92 Location receiving healthcare services (ref: Taipei region) Central Taiwan 1.07 0.93e1.24 1.15 0.97e1.37 Southern Taiwan 0.69y 0.58e0.83 0.80* 0.65e0.99 The Kaoping 0.65y 0.54e0.78 0.64y 0.52e0.80 region Northern Taiwan 0.73z 0.60e0.89 0.72z 0.57e0.91 Eastern Taiwan 0.70 0.48e1.03 0.72 0.46e1.13
1.31 1.83 -
0.99e1.72 y
1.38e2.43 -
-
-
0.92 0.49y 0.69*
0.71e1.19 0.34e0.68 0.49e0.98
0.78 0.64
0.55e1.10 0.31e1.33
Logistic regression with conditional forward method was used to explore the factors related to program adherence. The independent variables included gender, age 65 years, number of diabetes-related complications, CCI, COCI, time since diagnosis of diabetes, hospitalization in the previous 1 year, and location receiving healthcare services. *P < .05. y P < .0001. z P < .01
Our study findings were similar to the findings of Lee et al19 and Cheng et al.17 Healthcare utilization/expenses for patients in the P4P program increased in the initial period after enrollment. Long-term, beneficial effects of the P4P program might have been present if patients had adhered to the program at certain levels. The reported long-term, beneficial effects might imply that avoidable healthcare expenses can be diminished because patients’ conditions and possible problems are regularly monitored, detected, and managed through comprehensive follow-up/evaluations. However, further evidence related to the long-term effects of diabetes P4P programs on healthcare expenses and potential savings is still needed. The expenses of hospitalizations (T1eT3) for the adherence group were significantly less than those for the nonadherence group. This study finding implies that the total healthcare expense saving incurred in the adherence group might be related to the cost savings on the expense of hospitalizations. Further exploring the potential areas of significant cost savings (eg, the expense of hospitalizations) resulted from the P4P program is recommended. Factors Related to Program Adherence Few studies have been conducted to investigate factors related to P4P program adherence. Thus, we describe factors related to healthcare utilization/costs, and adherence to practice guidelines/dispensed medication among patients with diabetes for further discussions. Significantly lower healthcare utilization was observed in men compared with women in the numbers of physician/practitioner/ medical specialist visits and laboratory tests.35 Better continuity of care was significantly associated with fewer hospital/emergency visit36 and hospitalizations,8,37 and decreased healthcare costs.8,37 Renard et al38 found that (1) men were less likely to adhere to recommended guidelines, and (2) patients who had been admitted to the hospital in the past or current year better adhered to guidelines than did those who had not been admitted. Satman et al39 found that
patients with longer durations of diabetes had better guideline adherence. Analyses of Taiwan’s NHI claims data showed that patients in the diabetes P4P program or those treated by P4P-participating physicians were more likely to adhere to practice guidelines than those who were treated by non-P4P-participating physicians.20 Patients with longer durations of diabetes and higher continuity of care scores were more likely to have good medication adherence.8,40 Gender, continuity of care, and time since diagnosis of diabetes were identified as factors related to healthcare utilization/costs, and adherence to practice guidelines/dispensed medications.8,3540 In this study, relationships between program adherence and these factors were found. Findings presented in this study and previous studies suggest that healthcare professionals should pay special attention to patients who are male and newly diagnosed with diabetes, and who have low continuity of care scores. Further studies including interventions or educational strategies which could improve patients’ adherence and continuity of care level are suggested to achieve optimal effects, disease control and clinical outcomes. A previous study found that patients who had been admitted to the hospital possessed better practice guideline adherence.38 Gibson et al41 found that higher levels of medication adherence were associated with more physician visits and fewer inpatient admissions/ emergency visits. The study results of Gibson et al41 might explain the relationships between program adherence and diabetes-related hospitalization identified in this study. Nuti et al42 reported that no-shows to a primary care appointment were associated with increased risks of hospitalization among patients with diabetes who had recently been hospitalized. Further studies including management strategies that could improve patients’ adherence are suggested. Feasible strategies include (1) providing needed education/counseling, (2) increasing the convenience of provided care, (3) improving patients’ involvement, and (4) using reminders, reinforcement, or rewards.43 One study conducted in a rural area of Taiwan found that rural residents with diabetes tended to exhibit inadequate health-related behaviors.44 In this study, the mean COCI score for participants receiving healthcare services in the Taipei region (0.43) was significantly higher than those of participants receiving services in southern Taiwan (0.40), the Kaoping region (0.39), and eastern Taiwan (0.36). Better program adherence of participants receiving healthcare services in the Taipei region might be explained by participants living in an urban area43 and those with higher continuity of care scores40 being more likely to perform healthcare behaviors. There are several limitations of this study. First, we were unable to control for other potentially important variables (eg, ethnicity, education, cognitive status, and number/type of medication usage), although we did control for available confounders when we conducted the analyses. Second, this study could not analyze some outcome measures of diabetes care (eg, the hemoglobin A1C) because of a lack of information. Third, we used the administrative definition of adherence to the first-stage P4P program. It is possible that some cases may have been misclassified. Achieving the first-stage P4P program goal for patients who were not meeting glycemic goals might not be sufficient to be defined as delivering quality care.45 Fourth, the cohort is representative of the population of Taiwan. The ability to generalize our study results might be limited because of different settings and different designs of P4P programs. Fifth, improved outcomes with a decrease in healthcare spending cannot completely represent a high quality of care. Economic evaluations with controlled intervention designs are suggested to identify more precise estimations. Conclusions Our study results provide further evidence of the long-term, beneficial effects of the diabetes P4P program, although healthcare utilization/expenses cannot completely represent the delivery of
M.-J. Chi et al. / JAMDA xxx (2016) 1e7
quality care. Study results also showed that the total healthcare expense saving incurred in the adherence group might be related to the cost saving on the expense of hospitalizations. Further exploration of the long-term effects of diabetes P4P programs on healthcare expenses and the potential areas of significant healthcare cost savings is recommended. Gender, continuity of care, time since diagnosis of diabetes, hospitalization in the previous year, and location receiving healthcare services were identified as factors related to program adherence. Further studies including intervention, education, or management strategies, which can improve patients’ adherence and continuity of care levels, are suggested to achieve optimal effects, disease control, and clinical outcomes. Acknowledgments We greatly appreciate the individuals who provided information for our analyses and the financial support of College of Nursing, Taipei Medical University. References 1. Danaei G, Finucane MM, Lu Y, et al. National, regional, and global trends in fasting plasma glucose and diabetes prevalence since 1980: Systematic analysis of health examination surveys and epidemiological studies with 370 countryyears and 2.7 million participants. Lancet 2011;378:31e40. 2. Lozano R, Naghavi M, Foreman K, et al. Global and regional mortality from 235 causes of death for 20 age groups in 1990 and 2010: A systematic analysis for the Global Burden of Disease Study 2010. Lancet 2012;380:2095e2128. 3. Ministry of Health and Welfare. 2015 Causes of death statistics. Available at: http://www.mohw.gov.tw/cht/DOS/Statistic.aspx?f_list_no¼312&fod_list_no¼ 5012. Accessed October 25, 2015. 4. National Health Insurance Administration. National Health Insurance in Taiwan 2014-2015 annual report. Available at: http://www.nhi.gov.tw/ webdata/webdata.aspx?menu¼17&menu_id¼659&webdata_id¼2891&WD_ID¼ 897. Accessed October 25, 2015. 5. Jiang YD, Chang CH, Tai TY, et al. Incidence and prevalence rates of diabetes mellitus in Taiwan: Analysis of the 2000e2009 Nationwide Health Insurance database. J Formos Med Assoc 2012;111:599e604. 6. International Diabetes Federation. IDF diabetes atlas, 2013 (6th ed.). Available at: http://www.idf.org/diabetesatlas. Accessed October 25, 2015. 7. Chang K. Comorbidities, quality of life and patients’ willingness to pay for a cure for type two diabetes in Taiwan. Public Health 2010;124:284e294. 8. Cheng JS, Tsai WC, Lin CL, et al. Trend and factors associated with healthcare use and costs in type two diabetes mellitus: A decade experience of a universal health insurance program. Med Care 2015;53:116e124. 9. Chiou ST, Lin HD, Yu NC, et al. An initial assessment of the feasibility and effectiveness of implementing diabetes shared care system in Taiwan: Some experiences from I-Lan County. Diabetes Res Clin Pract 2001;54:S67eS73. 10. Rosenthal MB, Fernandopulle R, Song HR, Landon B. Paying for quality: Providers’ incentives for quality improvement. Health Aff 2004;23:127e141. 11. National Health Insurance Administration. 2011 National Health Insurance Pay-for-Performance program for diabetes (in Chinese). Available at: http:// www.nhi.gov.tw/webdata/webdata.aspx?menu¼20&menu_id¼712&webdata_ id¼3862&WD_ID¼836. Accessed October 25, 2015. 12. Chang RE, Lin SP, Aron DC. A pay-for-performance program in Taiwan improved care for some diabetes patients, but doctors may have excluded sicker ones. Health Aff 2012;31:93e102. 13. Chen PC, Lee YC, Kuo RN. Differences in patient reports on the quality of care in a diabetes pay-for-performance program between 1 year enrolled and newly enrolled patients. Int J Qual Health Care 2012;24:189e196. 14. Chen TW, Wang JY. Effectiveness of the diabetes shared care network for middle-aged and elderly patients in a rural area of Taiwan. Taiwan J Public Health 2009;28:334e343. 15. Lee CP, Chu CI, Wu DA. Comparison of the diabetes regular care model and the “Diabetes shared care disease management” model in eastern Taiwan. Tzu Chi Med J 2007;19:66e73. 16. Hsu HC, Weng HC, Lin YT, et al. A study of economic clinical and satisfaction evaluation of effectiveness of diabetes disease management program: A pilot study of a community hospital. J Health Manag (Taiwan) 2004;5:222e242. 17. Cheng SH, Lee TT, Chen CC. A longitudinal examination of a pay-forperformance program for diabetes care: Evidence from a natural experiment. Med Care 2012;50:109e116.
7
18. Hsieh HM, Tsai SL, Shin SJ, et al. Cost-effectiveness of diabetes pay-forperformance incentive designs. Med Care 2015;53:106e115. 19. Lee TT, Cheng SH, Chen CC, Lai MS. A pay-for-performance program for diabetes care in Taiwan: A preliminary assessment. Am J Manag Care 2010;16: 65e69. 20. Lai CL, Hou YH. The association of clinical guideline adherence and pay-forperformance among patients with diabetes. J Chinese Med Assoc 2013;76: 102e107. 21. Tan EC, Pwu RF, Chen DR, Yang MC. Is a diabetes pay-for-performance program cost-effective under the National Health Insurance in Taiwan? Qual Life Res 2014;23:687e696. 22. Huang J, Yin S, Lin Y, et al. Impact of pay-for-performance on management of diabetes: A systematic review. J Evid Based Med 2013;6:173e184. 23. Van Herck P, De Smedt D, Annemans L, et al. Systematic review: Effects, design choices, and context of pay-for-performance in health care. BMC Health Serv Res 2010;10:247. 24. Scott A, Sivey P, Ait Ouakrim D, et al. The effect of financial incentives on the quality of health care provided by primary care physicians. Cochrane Database Syst Rev 2011:CD008451. 25. Chew BH. Medication adherence on quality of life among adults with type 2 diabetes mellitus: An exploratory analysis on the EDDMQoL study. Qual Life Res 2015;24:2723e2731. 26. National Health Research Institutes (NHRI). 2014 National Health Insurance Research Database: Longitudinal cohort of diabetes patients. Available at: http://nhird.nhri.org.tw/date_lhdb.htm. Accessed July 15, 2014. 27. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: Development and validation. J Chronic Dis 1987;40:373e383. 28. Young BA, Lin E, Von Korff M, et al. Diabetes complications severity index and risk of mortality, hospitalization, and healthcare utilization. Am J Manag Care 2008;14:15e23. 29. Bice TW, Boxerman SB. A quantitative measure of continuity of care. Med Care 1977;15:347e349. 30. Central Bank of the Republic of China (Taiwan). 2015 Foreign exchange regime. Available at: http://www.cbc.gov.tw/content.asp?CuItem¼1879. Accessed September 20, 2015. 31. Ministry of Health and Welfare. Number of hospitals, clinics and beds by category and locale, 2008-2013. Available at: http://www.mohw.gov.tw/EN/ Ministry/Statistic.aspx?f_list_no¼474&fod_list_no¼5173. Accessed September 20, 2015. 32. Eijkenaar F, Emmert M, Scheppach M, Schöffski O. Effects of pay for performance in health care: A systematic review of systematic reviews. Health Policy 2013;110:115e130. 33. Fagan PJ, Schuster AB, Boyd C, et al. Chronic care improvement in primary care: Evaluation of an integrated pay-for-performance and practice-based care coordination program among elderly patients with diabetes. Health Serv Res J 2010;45:1763e1782. 34. Curtin K, Beckman H, Pankow G, et al. Return on investment in pay for performance: A diabetes case study. J Healthc Manag 2006;51:365e374. discussion 375e376. 35. Krämer HU, Rüter G, Schöttker B, et al. Gender differences in healthcare utilization of patients with diabetes. Am J Manag Care 2012;18:362e369. 36. Chen CC, Cheng SH. Better continuity of care reduces costs for diabetic patients. Am J Manag Care 2011;17:420e427. 37. Hong JS, Kang HC. Continuity of ambulatory care and health outcomes in adult patients with type 2 diabetes in Korea. Health Policy 2013;109: 158e165. 38. Renard LM, Bocquet V, Vidal-Trecan G, et al. Adherence to international followup guidelines in type two diabetes: A longitudinal cohort study in Luxembourg. PLoS One 2013;8:e80162. 39. Satman I, Imamoglu S, Yilmaz C, ADMIRE Study Group. A patient-based study on the adherence of physicians to guidelines for the management of type two diabetes in Turkey. Diabetes Res Clin Pract 2012;98:75e82. 40. Chang HY, Chiou CJ, Lin MC, et al. A population study of the self-care behaviors and their associated factors of diabetes in Taiwan: Results from the 2001 National Health Interview Survey in Taiwan. Prev Med 2005;40: 344e348. 41. Gibson TB, Song X, Alemayehu B, et al. Cost sharing, adherence, and health outcomes in patients with diabetes. Am J Manag Care 2010;16:589e600. 42. Nuti LA, Lawley M, Turkcan A, et al. No-shows to primary care appointments: Subsequent acute care utilization among diabetic patients. BMC Health Serv Res 2012;12:304. 43. Lerman I. Adherence to treatment: The key for avoiding long-term complications of diabetes. Arch Med Res 2005;36:300e306. 44. Chen MY, Huang WC, Peng YS, et al. Health status and health-related behaviors among type 2 diabetes community residents. J Nurs Res 2011;19:35e43. 45. American Diabetes Association. Standards of medical care in diabetes-2014. Diabetes Care 2014;37:S14eS80.