Effects of pay-for-performance system on
tuberculosis default cases control and treatment
in Taiwan
Wen-Chen Tsai
a, Pei-Tseng Kung
b, Mahmud Khan
c, Claudia Campbell
c,
Wen-Ta Yang
d, Tsuey-Fong Lee
e, Ya-Hsin Li
f,*
a
Department of Health Service Administration, China Medical University, Taichung, Taiwan, ROC
b
Graduate Institute of Health Administration, Asia University, Taichung, Taiwan, ROC
cDepartment of Health Systems Management, Tulane University, New Orleans, LA, USA d
Department of Internal Medicine, Taichung Hospital, Department of Health, Taichung, Taiwan
eCenters for Disease Control, R.O.C.(Taiwan), Third Branch Chief, Taichung, Taiwan f
Department of Health Policy and Management, Chung Shan Medical University and Hospital, 110, Sec 1, Jianguo N. Rd., 402 Taichung, Taiwan, ROC
Accepted 18 June 2010 Available online 4 August 2010
KEYWORDS
Pay-for-performance system;
Tuberculosis; Default rate
Summary Objectives: In order to make tuberculosis (TB) treatment more effective and to lower the default rate of the disease, the Bureau of National Health Insurance (BNHI) in Taiwan implemented the “pay-for-performance on Tuberculosis” program (P4P on TB) in 2004. The purpose of this study is to investigate the effectiveness of the P4P system in terms of default rate.
Methods: This is a retrospective study. National Health Insurance Research Datasets in Taiwan from 2002 to 2005 has been used for the study. The study compared the differences of TB de-fault rate before and after the implementation of P4P program, between participating and non-participating hospitals, and between P4P hospitals with and without case managers. Fur-thermore, logistic regression analysis was conducted to explore the related factors influencing TB patients default treatment after TB detected.
Results: The treatment default rate after “P4P on TB” was 11.37% compared with the 15.56% before “P4P on TB” implementation. The treatment default rate in P4P hospitals was 10.67% compared to 12.7% in non-P4P hospitals. In addition, the default rate was 10.4% in hospitals with case managers compared with 12.68% in hospitals without case managers.
* Corresponding author. Tel.:þ886 4 24730022x17172; fax: þ886 4 22412337. E-mail address:yli.tulane@gmail.com(Y.-H. Li).
0163-4453/$36ª 2010 The British Infection Society. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.jinf.2010.06.016
Conclusions: The results of the study showed that “P4P on TB” program improved the treat-ment default rate for TB patients. In addition, case managers improved the treattreat-ment out-come in controlling patients’ default rate.
ª 2010 The British Infection Society. Published by Elsevier Ltd. All rights reserved.
Background
TB has been a serious problem in Taiwan for years. The Department of Health (DOH) data indicates that the incidence rate is around 70 per 100,000 population in Taiwan and the fatality rate is around 4 per 100,000 population in year 2005. It appears that Taiwan has not been very successful in implementing effective TB control and management to achieve the international standard in case fatality rate. More rigorous TB control, treatment and management are required for achieving a better outcome in Taiwan.
In order to make TB treatment more effective and to avoid rapid spread of the disease, the Bureau of National Health Insurance (BNHI) in Taiwan started the implementa-tion of a pay-for-performance payment demonstraimplementa-tion project in October of 2001. Later in January 2004 the demonstration project was scaled-up as a national pro-gram. The program was officially named as “pay-for-performance on Tuberculosis” (P4P on TB). All diagnosis of “tuberculosis” (ICD-9-CM codes 010.0e018.9) in Taiwan was included in this program, but multi-drug resistant tuberculosis (MDR-TB) cases, extrapulmonary TB (EPTB) cases, and atypical mycobacterium tuberculosis cases were excluded. The BNHI set the objectives of pay-for-performance demonstration programs for TB to include: 1) improving the case management model; 2) decreasing the death rate from tuberculosis; 3) increasing the cure rate of tuberculosis; 4) improving the payment system to encour-age health care providers to provide high quality treat-ment, and give them more responsibilities.
The idea of pay-for-performance payment system de-rives from the concept of “case management”. Case management has been defined as “a collaborative process of assessment, planning, facilitation and advocacy for options and services to meet an individual’s health needs through communication and available resources to promote quality cost-effective outcomes,”.1Case management was developed for coordinating care so that continuous discov-ery and conformity can take place. Therefore, the P4P on TB program integrate the case managers into the system to provide continuous health care services to TB patients.
The most serious problem hampering tuberculosis ment and control is patients’ non-compliance with treat-ment or therapy.2e7 Defaulting or lack of adherence with the treatment regime increase the transmission of disease within the community, increase the morbidity rate and in-crease the cost of TB control efforts.3,6
Several countries have implemented many incentives activities on Tuberculosis control and treatment in recent years in which the ideas were consistent with pay-for-performance. The incentives can be to patients or to providers. For patients, most countries provide food or other living supports as incentives to encourage patients. In Russian, some cities provided food, travel support or
clothing to patients if they did not interrupt treatment. The default rates dropped from 15e20% to 2e6%. In Tajikistan, vulnerable patients were given food, conditional on their adherence to treatment. The treatment success rate was 89.5% compared to 59.4% in non-incentive pa-tients.8From the provider incentive side, several countries also showed positive results. In China, providers would re-ceive financial payment or monetary incentives when diag-nosis or treatments have been completed. However, no related results are showed. Bangladesh provided travel, food and doctor handbag to TB treatment doctors, and the cure rate increased significantly to 88%.9
The first study which examined the effectiveness of pay-for-performance demonstration project on TB in Taiwan was conducted in 2003.10 The results suggested that the project can shorten the period of tuberculosis treatment for the patients cured. Same group11 did fur-ther study and indicated that TB cure rate and length of treatment for cured cases improved significantly after P4P on TB implemented. Hospitals participating in the program showed significant better cure rate and length of treatment for cured cases comparing with non-partici-pating hospitals.
Although the pay-for-performance has been imple-mented in Taiwan over several years, only a few studies discussed the effectiveness of the policy.10e18Most studies on tuberculosis focused on the epidemiology or public health aspects of the disease. No related studies were fo-cused on the TB interrupt problems. This study will focus on the outcomes of the incentive structure of the P4P on TB program by analyzing the data set obtained from the na-tional insurance agency. More specifically, the study will compare the default rate of TB treatment before and after the implementation of P4P on TB, between hospitals by participation status in the P4P program. Another related comparison will be to see if participating hospitals with case managers perform better on patients’ interruption than the hospitals without case managers.
Research methods
Data sources
National Health Insurance Research Dataset in Taiwan from 2002 to 2005 has been used for the study. Total study samples are 13,191 patients in the years 2002 and 2003, and 27,142 patients in the years 2004 and 2005. For the analysis of pay-for-performance program on TB default rate, additional data were also obtained from the National Health Insurance Research Dataset. These additional variables include pa-tients’ gender, age, area of residence, papa-tients’ ICD-9-CM co-des, levels of comorbidity, notification physician, date of medical visits, date of inpatient/discharge, duration of drug therapy, notification hospital/institution, etc.
Inclusion and exclusion criteria
The Department of Health and WHO19defined the “default rate” as the percent of patients who interrupt the treat-ment regime for 2 months or longer. The duration of default is calculated by taking the difference between a medical visit and the next visit when the gap is 2 months or more. The date of medical visits and duration of medication days provide information to understand the patients’ com-pliance behavior. Patient with as long as 4 months of treat-ment from start of treattreat-ment was included in the study since diagnosis errors are common in the beginning of TB treatment.
The P4P on TB demonstration program was implemented in the central branch of TCDC (Center for Disease Control, Taiwan) in October, 2001. For the purpose of this analysis, the cases from the central branch of TCDC were excluded from the list of pre-P4P cases. Another 20,167 patients were excluded from the study because of diagnosis error in the early stage of treatment.
Data analysis
Descriptive statistics were obtained for the number of default cases and default rate for TB patients for the years 2002e2005. Besides descriptive analysis, bivariate analysis
was performed to understand the effects of “P4P on TB” program on the default rate or treatment interruption. Chi-square test and fishers’ exact test were used to compare patients’ default rates between pre-P4P and post-P4P, be-tween P4P hospital and non-P4P hospital, and bebe-tween P4P hospitals with and without case managers.
To identify the factors affecting interruption from treatment in Taiwan, multivariate analysis was conducted. Since the outcome variable is dichotomous, the appropriate empirical model would be to use logistic regression equa-tions. Logistic regression analyses were performed to estimate the probability of interruption from treatment. The independent variables included in the analysis were gender, age, migration status, hospital characteristics, comorbidity level, levels of city development, city income, and physician resources.
Level of comorbidity was followed by the Charlson comorbidity index (CCI) which was developed by Charlson et al. in 1987.20 Deyo et al. adapted the CCI for use with CM administrative data by identifying specific ICD-9-CM code to represent each of the individual comorbid con-ditions in the original Charlson index. This study used this index to identify the levels of patients’ comorbidity. The higher the CCI points the more serious is the patients’ co-morbid condition. Three CCI classes (no, mild, and severe) were defined as 0, 1, and2.
Table 1 Comparison of TB patients’ default rate: before and after the implementation of P4P.
Before P4P P After P4P P Pa
No default Default No default Default
N (%) N (%) N (%) N (%) N 11,138 84.44 2053 15.56 24,057 88.63 3085 11.37 <0.05 Gender 0.01 <0.05 0.31 Male 7465 83.85 1438 16.15 16,148 87.92 2218 12.08 Female 3555 85.62 597 14.38 7872 90.11 864 9.89 Age (years) <0.05 <0.05 0.60 <45 3089 86.53 481 13.47 6610 89.72 757 10.28 45e65 3136 83.67 612 16.33 6869 88.53 890 11.47 >65 4913 83.65 960 16.35 10,578 88.03 1438 11.97 Migration <0.05 <0.05 0.26 No 10,906 85.16 1900 14.84 23,701 89.34 2828 10.66 Yes 232 60.26 153 39.74 356 58.08 257 41.92 Comorbidity levels <0.05 <0.05 0.00 No 10,548 85.00 1862 15.00 22,947 88.99 2839 11.01 Mild 561 74.90 188 25.10 1021 81.81 227 18.19 Severe 29 90.63 3 9.38 89 82.41 19 17.59
City income levels <0.05 <0.05 0.01
High 2673 82.42 570 17.58 5777 88.55 747 11.45
Median 8004 85.47 1361 14.53 17,361 88.83 2182 11.17
Low 447 79.68 114 20.32 897 85.51 152 14.49
Physician density levels <0.05 <0.05 0.04
High 2559 82.07 559 17.93 5573 88.21 745 11.79
Median 7282 85.68 1217 14.32 15,511 89.07 1903 10.93
Low 1283 82.67 269 17.33 2951 87.20 433 12.80
Result
From the data sets of Bureau of National Health Insurance, the number of new TB cases was 13,191 in 2002 and 2003 (pre-P4P period), and 27,142 in 2004 and 2005 (post-P4P period). The data set implies that 15.56% (NZ 2053) of patients interrup-ted treatment schedule for more than 2 months during the pre-P4P study period while treatment default rate was 11.37% (NZ 3085) during post-P4P years. The default rate de-creased significantly after “P4P on TB” program was imple-mented (P< 0.05).Table 1indicates that after “P4P on TB” implemented, the default rates significantly decreased among patients after controlling for comorbidity levels, levels of city income, and physician resources as compared with the default rate before (P< 0.05).
The default rates based on patient characteristics and city characteristics were almost identical between pre and post-P4P scenarios. As indicated byTable 1, treatment de-faults were significantly different by patients’ gender, age groups, migration status, comorbidity levels, levels of city income, and levels of physician resources (P< 0.05). Male had significantly higher default rate than females and youn-ger patients had significantly lower default rate than older patients. Migrating patients had significantly higher default rate than patients who are not migrants (P< 0.05). Patients
in cities with the lowest income (lower than mean of in-come in all cities in Taiwan minus one S.D) levels show the highest default rate (20.32%, 14.49%) while cities with median income level (mean of income in all cities in Taiwan one S.D) had the lowest default rate (14.53%, 11.17%). Patients from cities with median levels of physi-cian resources (physician resource of median in Taiwan one S.D) show the lowest default rate (14.32%, 10.93%).
To compare the default rates between P4P and non-P4P hospitals, the study focused on the medical centers and regional hospitals in Taiwan after the implementation of “P4P on TB”. Seventeen out of 24 medical centers and 57 out of 78 regional hospitals joined the “P4P on TB” program in the year 2004. In the years 2004 and 2005, a total 638 TB cases were found in non-P4P hospitals whereas 16,423 TB cases were found in P4P hospitals.
As indicated in Table 2, the default rate was lower in hospitals participating in P4P compared to non-participat-ing hospitals (P< 0.05). The default rate was 10.67% in hos-pitals participating in P4P compared to 12.7% in non-participating hospitals. The default rates were significantly lower in P4P hospitals than non-P4P hospitals after control-ling for gender, age, migration status, comorbidity level, levels of city income, and physician resources compared with non-P4P hospitals (P< 0.05).
Table 2 Comparison of TB patients’ default rate between P4P hospitals and non-P4P hospitals.
P4P hospitals P Non-P4P hospitals P Pa
No default Default No default Default
N (%) N (%) N (%) N (%) N 14,671 89.93 1752 10.67 557 87.30 81 12.70 <0.05 Gender 0.01 0.03 <0.05 Male 9765 88.91 1218 11.09 352 85.23 61 14.77 Female 4892 90.21 531 9.79 204 91.07 20 8.93 Age (years) 0.20 0.39 0.04 <45 4039 89.84 457 10.16 172 88.66 22 11.34 45e65 4158 89.59 483 10.41 171 84.65 31 15.35 >65 6474 88.86 812 11.14 214 88.43 28 11.57 Migration <0.05 0.02 <0.05 No 14,630 89.62 1694 10.38 546 87.78 76 12.22 Yes 41 41.41 58 58.59 11 68.75 5 31.25 Comorbidity levels <0.05 0.42 <0.05 No 13,854 89.76 1580 10.24 520 87.54 74 12.46 Mild 747 82.91 154 17.09 32 82.05 7 17.95 Severe 70 79.55 18 20.45 5 100.00 0 0.00
City income levels <0.05 0.46 <0.05
High 3851 87.82 534 12.18 53 82.81 11 17.19
Median 10,482 89.92 1175 10.08 482 87.96 66 12.04
Low 324 89.26 39 10.74 22 84.62 4 15.38
Physician number levels <0.05 0.78 <0.05
High 3661 87.35 530 12.65 337 87.99 46 12.01
Median 8849 90.89 887 9.11 204 86.44 32 13.56
Low 2147 86.64 331 13.36 16 84.21 3 15.79
In respective of the participation in P4P, males had higher default rate than females, patients migrating from one place to another had much higher default rate than the non-migrant patients (P< 0.05). In P4P hospitals, patients with no comorbidity (CCIZ 0) had the lowest default rate (10.24%) whereas patients with severe comorbidity levels (CCIS 2) had the highest default rate (20.45%) (P < 0.05). For P4P hospitals, patients from cities with highest income level had the highest default rate (12.18%) while patients from cities with median levels of income had the lowest rate (10.08%). If we examine the default rates by physician resources, cities with the lowest physician resources showed the highest default rates for both the P4P and non-P4P hospitals (seeTable 2).
The study also examined the difference in rates by hospitals with case managers and hospitals without case managers. The study had 1900 new TB cases from hospitals without case managers and 14,523 new cases from hospitals with case managers. Table 3 summarizes the descriptive statistics for default rate by hospitals with or without case managers. The table shows that the default rate in hospitals with case managers was 10.4%, which is signifi-cantly lower than the rate in hospitals without case man-agers (12.68%) (P< 0.05). The default rates in hospitals with case managers were also lower for each of the
population categories defined by controlling for patients’ gender, migration status, comorbidity levels, levels of city income, and of physician resources (P< 0.05) (seeTable 3). Therefore, since the default rates are lower for each of the patient and city characteristics for hospitals with case managers, presence of case managers clearly had a positive impact on treatment contribution. Patients with no comor-bidity level (CCIZ 0) had lower default rates than patients with median (CCIZ 1) or severe comorbidity (CCI S 2) level. Patients having moved to other places had signifi-cantly higher default rates than patients without migration history (P< 0.05). Patients living in cities with the highest physician resources had the lowest default rates while pa-tients live in cities with median level of physician resources had the lowest default rates.
In the hospitals with case managers, male patients had significantly higher default rates (10.86%) than the female patients (9.46%) (P< 0.05), and younger patients had lower default rates (9.74%) than the older patients (P< 0.05). Pa-tients living in the cities with the highest income level had the highest default rates (12.13%), followed by cities with the lowest income levels (10.43%) and cities with median levels of income (9.75%) (Table 3).
Table 4 reports the results of the logistic regression model with treatment interruption probability for TB
Table 3 Comparison of TB patients’ default rate between P4P hospitals with case managers and P4P hospitals without case managers.
P4P hospitals with case managers P P4P Hospitals without case managers P Pa
No default Default No default Default
N (%) N (%) N (%) N (%) N 13,012 89.60 1511 10.40 1659 87.32 241 12.68 <0.05 Gender 0.01 0.84 0.01 Male 8635 89.14 1052 10.86 1130 87.19 166 12.81 Female 4366 90.54 456 9.46 526 87.52 75 12.48 Age (years) 0.17 0.67 0.69 <45 3522 90.26 380 9.74 517 87.04 77 12.96 45e65 3735 89.72 428 10.28 423 88.49 55 11.51 >65 5755 89.11 703 10.89 719 86.84 109 13.61 Migration <0.05 <0.05 <0.05 No 12,979 89.86 1464 10.14 1651 87.77 230 12.23 Yes 33 41.25 47 58.75 8 42.11 11 57.89 Comorbidity level <0.05 0.03<0.05 No 12,312 90.01 1367 9.99 1542 87.86 213 12.14 Mild 638 83.40 127 16.60 109 80.15 27 19.85 Severe 62 78.48 17 21.52 8 88.89 1 11.11
Levels of city income <0.05 0.87<0.05
High 3348 87.87 462 12.13 503 87.48 72 12.52
Median 9342 90.25 1009 9.75 1140 87.29 166 12.71
Low 309 89.57 36 10.43 15 83.33 3 16.67
Levels of physician resources <0.05 <0.05 <0.05
High 3123 88.12 421 11.88 538 83.15 109 16.85
Median 7938 90.88 797 9.12 911 91.01 90 8.99
Low 1938 87.02 289 12.98 209 83.27 42 16.73
patients as the dependent variable. The independent vari-ables included in the model are gender, age, migration sta-tus, hospital participating in P4P program, hospital with/ without case managers, comorbidity level, levels of city in-come, and physician resources. Note from the parameters estimated (Table 4) that the probability of default is lower for female compared to male patients (ORZ 0.878). Pa-tients migrating to other cities were more likely to inter-rupt their treatment than patients not migrating during the treatment phase (ORZ 9.829). As expected, patients with more serious comorbidity status are more likely to in-terrupt from treatment than patients with no comorbidity. Patients with mild comorbidity level (CCIZ 1) had around 1.6 times more likely to default from treatment while pa-tients with severe comorbidity level (CCIS 2) had around 2 times more likely to default compared to patients with no comorbidity. Patients in P4P hospitals were less likely to default than patients in non-P4P hospitals (ORZ 0.904), and patients in hospitals with case manager were less likely to default than patients in hospitals without case manager (ORZ 0.831). A number of city characteris-tics also influenced the default rate. Patients living in cities
with median level of income were less likely to default from treatment than patients living in cities with high income level (ORZ 0.881). Patients in cities with median level of physician resources were less likely to default than patients living in cities with high income levels (ORZ 0.757) (see Table 4).
After “P4P on TB” program was implemented, the study found total 3085 patients with 6022 times of default from treatment. There was 15.84% (NZ 954) of the defaults within the first 3 months of treatment, and 30.78% (NZ 1854) defaulted between the fourth and sixth month of treatment, and 29.26% (NZ 1762) defaulted between the seventh and ninth month of treatment, and 24.12% (NZ 1452) defaulted after the tenth month of treatment. The number of default cases dramatically increased from the first month to the 3rd month of treatment, and peaked at the 5th month from the start of treatment (Fig. 1). After the 5th month, the trend of number of default cases de-creased slightly. Furthermore, the study analyzed the rela-tionship between time from start of treatment to default and patients’ medication days per medical visits. Patients receiving medication days between 15 to 28 days had the Table 4 Results of logistic regression analysis with TB cases default from treatment more than 2 months as the dependent variable.
Criteria Estimate OR (95% CI) P-value
Intercept 1.8806 <.0001 Gender Male (reference) Female 0.1303 0.878 (0.788e0.978) 0.0183 Age (years) <45 (reference) 45-65 0.0458 1.047 (0.915e1.198) 0.5048 >65 0.0951 1.100 (0.972e1.244) 0.1305 Migration No (reference) Yes 2.2853 9.829 (6.757e14.298) <.0001 Comorbidity level No (reference) Mild 0.4716 1.603 (1.335e1.924) <.0001 Severe 0.6905 1.995 (1.187e3.353) 0.0092 P4P Hospital No (reference) Yes 0.1011 0.904 (0.702e0.973) 0.0241 Case manager No (reference) Yes 0.1849 0.831 (0.717e0.964) 0.0142
Levels of city income Low (reference) Median 0.1267 0.881 (0.768e1.011) 0.0710 High 0.1060 1.112 (0.784e1.576) 0.5518 Physician resources Low (reference) Median 0.2785 0.757 (0.658e0.871) 0.0001 High 0.1257 1.134 (0.954e1.347) 0.1529
highest default rate which was 35.77%, following by the pa-tients receiving medication days less than 7 days (24.48%). The lowest default rate was in the group of patients with medication days more than 28 days (Fig. 2).
Discussions
The study concluded that P4P on TB program could de-crease the default rate significantly in Taiwan. Although some other countries implemented related incentive pro-grams on TB control, the related outcomes are still not available. However, comparing with basic report from Russian, the incentive program significantly decreased the default rate.8 Russian focused on the patients’ incentive, which provided patients some food or living supports to in-crease patients’ compliance.9 Government in Taiwan can consider the patients’ incentive activities. For poor pa-tients, food or traveling support may be needed. For non-poor patients, some extra monetary incentives may en-courage patients’ compliance and decrease the default rate.
From this study, male had lower compliance (higher default rate) than women. Some previous studies had same conclusions.3,10,21e23 Previous study found that once the symptoms were gone, men and the elderly were easily de-faulted from treatment.2 Men and elderly do not under-stand the importance of compliance and completing the full treatment. Same findings were in the Philippines be-cause men were too hard to give up alcohol. Drinking plays
an important role makes a man as seriously ill.24Therefore, the patients’ education especially for man is important for improving their compliance behaviors.
Age had effects on univariate analysis. However, after controlling for other variables, age became a non-signifi-cant factor. The study conducted the relationship between age and comorbidity level, and found that elderly patients had severe comorbidity than non-elderly (P< 0.001). In the age group of older than 65 years old, 91% of patients with-out any comorbidity compared with 94.5% and 97.2% in aged 45e65 years old and less than 45 years old. Therefore, the age factor becoming insignificant may be due to influ-ence of comorbidity level in elderly patients.
Stigma of TB was the other main reason for patients default from treatment.2 Previous studies also concluded same results that stigma is the main reason for patient in-terrupting treatment.4,25,26Because of the stigma attached to TB, patients usually hide their condition from their em-ployers or friends.2,4 From clinical experience, the TB treatment team in Taiwan found that non-compliance due to stigma may be described as more common among men in Taiwan because men were often the only earning mem-ber in a family. Several previous studies also mentioned that lacking of social support was the main problem of pa-tient default from treatment, especially when they felt better soon after starting treatment.27,28Health education for patients and society is the only way to improve the problem of stigma.
The study indicates that older patients had tendency of higher default rate than younger patients although it is not significant. Some studies had similar outcome with our study.22,24This may be due to the fact that the side-effects of many drugs are more frequent or more severe in elderly patients.24 In addition, older patients who live in rural areas may have traveling problems when they seek for med-ical services. Since the directly observed treatment, short-course (DOTS) program has been implemented in Taiwan in 2006, the traveling problems for these patients might be improved.
Another interesting part for the result showed the default rate for migrants in 2002 and 2003 (39.74%) was lower than default rate for migration patients in 2004 and 2005 (41.92%). That might be the severe acute respiratory syndrome (SARS) outbreak in Taiwan rapidly in 2003. In order to control the infectious of SARS efficiently, the migrants decreased rapidly. In addition, since SARS ex-plored in the society rapidly, patients’ compliance behav-iors might be improved because of the awareness of disease control. Therefore, that might decrease the default problems during this period.
Consistent with the findings in previous studies,5,29,30 migrants had the worst outcomes in terms of completion of treatment. Europe and USA have been aware of this problem for several years.29e33Changes of domicile will af-fect patients’ behaviors.34,35 Lack of fixed address is the main problem for migrants. As the results of comparing the default rate in 2004 and 2005, P4P hospitals or hospitals with case managers did not show lower default rate for mi-gration of individuals. However, when the study only fo-cused on the default cases, the migration cases were 3.31% and 3.11% of P4P hospitals and with case managers, respectively. Theses numbers were lower than non-P4P
0 2 4 6 8 10 12 14 0 1 2 3 4 5 6 7 8 9 10 11 12
Treatment duration in months
Percent of interrupted cases (%)
Figure 1 Percent of cases interrupting treatment by treat-ment duration.
Figure 2 The relationship between default rate and pre-scription days.
hospital and without case managers (6.17% and 4.56%, re-spectively). This revealed the function of case managers who monitor and update patients’ information should be more strengthened. Once patients move to other cities, the case managers should transfer patients’ information to next charged case managers. Therefore, the training of case managers should be improved and the communications between hospital and community should be also enhanced. The multivariate analysis presents the default rate in the hospitals participating in “P4P on TB” is significantly lower than non-participating hospitals. In addition, partici-pating hospitals with case managers had significantly lower default rate than hospital without case managers. The results confirmed that presence of case managers clearly had a positive impact on TB treatment contribution. In addition, the results also achieve one of the objects of “P4P on TB” program which improving case management system in TB treatment.
In the respect of levels of physician resources, the study found that patients living in the highest level of physician resources did not have the lowest default rate. Serious patients might be transferred from hospitals in rural areas to hospitals in urban cities which have more health care resources in cities.36Traveling time was also one of the de-termined factors associated with default.23Patients might default from routine outpatient visits due to the long dura-tion of transportadura-tion. Since the DOTS program in Taiwan has been implemented in 2006, the problem of transporta-tion might be decreased rapidly.
The finding in this study suggested that default could be associated with the time from start of treatment. As our studies’ results, at the 5th month from start of treat-ment was the month with the highest default rate. The study from Singapore had similar findings in year 2000.37 Many studies concluded that higher compliance of TB treatment could be attributed to the presence of disease symptoms, because compliance is usually lower with treatment for asymptomatic conditions.38,39 For the treatment of tuberculosis, the disease symptoms were usually resolved within the first 3 months of treatment. Thus patients may discontinue the treatment when they feel markedly better after the first 3 months of treat-ment.28,40 This underscores the importance of TB pa-tients’ education on the need to complete the full course of treatment so as to prevent subsequent relapse. In addition, the ultimate responsibility for this can only lie on the treating physicians and case managers, whose role was not only to treat disease, but also to ensure that the patients adhere to and complete the treatment. Previous study suggested that reminding patients by clinics or hospitals are required to ensure the rate of com-plete treatment.41 Telephone reminders, use of incen-tives are effective methods.
No related studies discussed about the relationship between medication days and default rate. According to the present study’s results, patients with no medication days or patients with medication days between 15 and 28 days had the highest default rate. According to the treatment guidelines for TB treatment, patients will get longer medication days when the disease symptoms de-clared and became stable. Patients might think they fully recovered from an illness and stopped the medication.
Again, case managers and physicians have the responsibility to educate patients’ behaviors. The studies’ results also found that the medication days would be associated with the time from start of treatment to default. The case man-agers should refer to the default time under different pa-tients’ medication days, and remind patients for medication visits following by the schedules.
There are some limitations for this study, the study just had limited variables for analysis from the data set which obtained from BNHI of Taiwan. The study also could not get sufficient hospital records to review other influencing factors such as the microbiologic data, drug susceptibility tests results and resistance pattern, presence of adverse reactions of anti-tubercular therapy and the comorbid illness of the patients.
Conclusions
In short, from the comparisons of “P4P hospitals” and “non-P4P hospitals”, the study found that the default rate was improved significantly. These results confirmed that P4P on TB is a good payment policy for TB controlling and treatment on patients’ default rate. The study also found the default rate in the hospitals with case managers was significantly lower than hospitals without case managers. The results showed that the case managers improved the problem of interruption.
Acknowledgements
We would like to thank Taiwan’s National Science Council (grant NSC95-2314-B-039-026) and China Medical University (grant CMU97-178) for providing financial support for this study. We are also grateful to the Centers for Disease Con-trol of Taiwan for providing data on TB patients.
References
1. Daniels Stefani, Ramey Marianne. The leader’s guide to hospi-tal case management. Jones & Bartlett Publishers; 2005. 2. Johansson E, Diwan VK, Huong ND, Ahlberg BM. Staff and
pa-tient attitudes to tuberculosis and compliance with treatment: an exploratory study in a district in Vietnam. Tubercle and Lung Disease 1996;77(2):178e83.
3. Al-Hajjaj MS, Al-Khatim IM. High rate of non-compliance with anti-tuberculosis treatment despite a retrieval system: a call for implementation of directly observed therapy in Saudi Ara-bia. The International Journal of Tuberculosis and Lung Dis-ease 2000;4(4):345e9.
4. Liam CK, Lim KH, Wong CMM, Tang BG. Attitudes and knowl-edge of newly diagnosed tuberculosis patients regarding the disease, and factors affecting treatment compliance. The In-ternational Journal of Tuberculosis and Lung Disease 1999;3: 300e9.
5. Jaggarajamma K, Sudha G, Chandrasekaran V, Nirupa K, Thomas A, Santha T, et al. Reasons for non-compliance among patients treated under revise3d national tuberculosis control programme (RNTCP), Tiruvallur district, South India. Indian Journal of Tuberculosis 2007;54:130e5.
6. Barnhoorn F, Adriaanse H. In search of factors responsible for non compliance among tuberculosis patients in Wardha Dis-trict, India. Social Science & Medicine 1992;34:291e306.
7. Brudney K, Dobkin J. Resurgent tuberculosis in New York City. Human immunodeficiency virus, homelessness, and the decline of tuberculosis control programs. The American Review of Re-spiratory Disease 1991;144(4):745e9.
8. Beith A, Eichler R, Weil D. Worldwide: incentives for tuberculosis diagnosis and treatment. Center for Global Development; 2009. 9. Eichler Rena. Can “Pay for Performance” increase utilization by
the poor and improve the quality of health services? Discussion paper for the first meeting of the working group on perfor-mance-based incentives. Center for Global Development; 2006. 10. Tsai WC, Kung PT, Chiang TD. Evaluating the effectiveness of pay-for-performance demonstration program on tuberculosis. Grant of the Bureau of National Health Insurance, Taipei, Tai-wan; 2003.
11. Tsai WC, Kung PT, Shih SM, Shen KH, Wu YC. The effects of pay-for-performance demonstration program on tuberculosis on tu-berculosis prevention. Grant of the Bureau of National Health Insurance, Taipei, Taiwan; 2005.
12. Lee WP, Sheu ML, Wu YC, Tang CH, Huang KC, Juang JJ. Anal-ysis of tuberculosis case management models in hospitals. Tai-wan Journal of Public Health 2008;27(5):373e84.
13. Tsai YC, Tai FT, Lee WH, Kao TT. The effectiveness of type 2 diabetic patient management in the improvement program of national health insurance payment for diabetes medical treat-ment. Show Chwan Medical Journal 2008;8(3e4):93e9. 14. Tsai WC, Kung PT. Factors associated with the tuberculosis
pa-tients’ failed rate. Department of Health, Taiwan, R.O.C.: Grant of the Center for Disease Control; 2003.
15. Tsai WC, Kung PT, Chiang ZD. Evaluation of NHI medical care improvement program fro tuberculosis patients. Grant Report of the Bureau of National Health Insurance; 2002.
16. Hsu HC, Weng HC, Lin YT, Chen SM, Lee PJ, Lee JM, et al. A study of economic, clinical and satisfaction evaluation of the effectiveness of diabetes disease management program e a pi-lot study of a community hospital. Journal of Healthcare Man-agement 2004;5(2):222e42.
17. Chen MJ, Chen CJ, Chien CW. Assessing the potential effects of the pay-by-quality system on outpatient expenses of diabetic patients. Journal of Healthcare Management 2003;4(4):14e26. 18. Lee YM, Chou CH, Weng HC, Yeh CR, Lee YC, Hsu HC, et al. The effectiveness of pay-for-performance on asthma. Taiwan Med-ical Journal; 2004.
19. World Health Organization. Treatment of tuberculosis: guide-lines for national programmes. Geneva, Switzerland: WHO Global Tuberculosis Program; 1997.
20. Charlson M, Pompei P, Ales K, MacKenzie C. A new method of classifying prognostic comorbidity in longitudinal studies: de-velopment and validation. Journal of Chronic Disease 1987; 40:373e83.
21. Dodor EA, Afenyadu GY. Factors associated with tuberculosis treatment default and completion at the Effia-Nkwanta Re-gional Hospital in Ghana. Transactions of the Royal Society of Tropical Medicine and Hygiene 2005;99(11):827e32. 22. Dodor EA. Tuberculosis treatment default at the communicable
diseases unit of Effia-Nkwanta Regional Hospital: a 2-year ex-perience. The International Journal of Tuberculosis and Lung Disease 2004;8:1337e41.
23. Ngamvithayapong-Yanai J, Puangrassami P, Yanai H. Compli-ance to tuberculosis treatment: a gender perspective. Gender and tuberculosis. Go¨teborg, Sweden: Nordic School of Public Health; 1998. p. 127e48.
24. Nichter N. Illness semantics and international health: the weak lungs/TB complex in the Philippines. Social Science & Medicine 1994;38(5):247e52.
25. Johansson E, Long NH, Diwan VK, Winkvist A. Attitude to com-pliance with tuberculosis treatment among women and men in Vietnam. The International Journal of Tuberculosis and Lung Disease 1999;3:862e8.
26. Demissie M, Getahun H, Lindtjørn B. Community tuberculosis care through “TB clubs” in rural North Ethiopia. Social Science & Medicine 2003;56(10):2009e18.
27. Khan Amir, Walley John, Newell James, Imdad Naghma. Tuber-culosis in Pakistan: socio-cultural constraints and opportunities in treatment. Social Science & Medicine 2000;50(2):247e54. 28. Shargie EB, Lindtjørn B. Determinants of treatment adherence
among smear-positive pulmonary tuberculosis patients in Southern Ethiopia. PLoS Medicine 2007;4(2):e37.
29. Comolet TM, Rakotomalala R, Rajaonarioa H. Factors deter-mining compliance with tuberculosis treatment in an urban en-vironment, Tamatave, Madagascar. International Journal of Tuberculosis and Lung Disease 1998;2(11):891e7.
30. Teale C, Goldman JM, Pearson SB. The association of age with the presentation and outcome of tuberculosis: a five-year sur-vey. Age and Ageing 1993;22:289e93.
31. Cayla` JA, Caminero JA, Rey R, Lara N, Valle`s X, Galdo`s-Tangu¨is. Current status of treatment completion and fatality among tu-berculosis patients in Spain. International Journal of Tubercu-losis & Lung Disease 2004;8(4):458e64.
32. Davies PDO. Tuberculosis and migration. In: Davies PDO, editor. Clinical tuberculosis. 2nd ed. London: Chapman and Hall; 1998. p. 365e82.
33. Rieder HL, Zellweger JP, Raviglione MC, Keizer ST, Migliori GB. Tuberculosis control in Europe and international migration. The European Respiratory Journal 1994;7:1545e53.
34. Colledge M, van Genus HA, Svensson PG, editors. Migration and health: towards an understanding of the health care needs of ethnic minorities. Proceedings of a consultative group on ethnic minorities, The Hague, The Netherlands, 28e30 November 1983. World Health Organization Regional Office for Europe; 1983. 35. World Health Organization Regional Office for Europe.
Tuber-culosis among migrant worker. Report on a working group, Bern 22e24 April 1975.
36. Sumartojo E. When tuberculosis treatment fails, a social be-havioral account of patients adherence. The American Review of Respiratory Disease 1998;147:1311e20.
37. Chee CBE, Boudville IC, Chan SP, Zee YK, Wang YT. Patient and disease characteristics, and outcome of treatment defaulters from the Singapore TB control unitda one-year retrospective survey. The International Journal of Tuberculosis and Lung Dis-ease 2000;4(6):496e503.
38. Cuneo WD, Snider DE. Enhancing patient compliance with tuber-culosis therapy. Clinics in Chest Medicine 1989;10(3):375e80. 39. Sbarbaro JA. Compliance: inducements and enforcements.
Chest 1979;6(Suppl.):750e6.
40. Jaiswal A, Singh V, Ogden JA, Porter JD, Sharma PP, Sarin R, et al. Adherence to tuberculosis treatment: lessons from the urban setting of Delhi, India. Tropical Medicine & International Health 2003;8(7):625e33.
41. Parsyan AE, Saukkonen J, Barry MA, Sharnprapai S, Horsburgh Jr CR. Predictors of failure to complete treatment for latent tuberculosis infection. Journal of Infection 2007; 54(3):262e6.
