- Split View
-
Views
-
Cite
Cite
Tsung-Tai Chen, Ya-Seng (Arthur) Hsueh, Chun-Hsiung Ko, Ling-Na Shih, Sien-Sing Yang, The effect of a hepatitis pay-for-performance program on outcomes of patients undergoing antiviral therapy, European Journal of Public Health, Volume 27, Issue 6, December 2017, Pages 955–960, https://doi.org/10.1093/eurpub/ckx114
- Share Icon Share
Abstract
To examine the effect of a participatory pay-for-performance (P4P) program in Taiwan on health outcomes for patients with severe hepatitis B or C.
This study adopted 4-year panel data from the databases of the National Health Insurance Administration (NHIA) in Taiwan. Using the caliper matching method to match patients in the P4P (experimental) group with those in the potential comparison group on a one-to-one basis for the year 2010, we tracked patients up to the year 2013 and employed Cox proportional-hazards regression models to evaluate the effect on patient outcomes.
The P4P group did not have a lower risk (HR = 0.44, P = 0.05) of hospital admission for severe hepatitis patients (i.e. need antiviral therapy). The risk of developing liver cirrhosis was also lower, but the reduction was not statistically significant (HR = 0.92, P = 0.77).
This study found that participatory-type P4P has not resulted in reduced hospital admission of hepatitis B or C patients who need antiviral therapy. The means by which the participatory P4P program could strengthen patient-centered care to achieve better patient health outcomes is discussed in detail.
Introduction
Most health care programs adopting pay-for-performance (P4P) systems are provider-focused. The systems are typically designed to provide extra incentives and establish a fee for service or capitation to meet objectives that have been agreed upon beforehand, such as the P4P systems adopted by the US and UK.1,2 Although the effects of many of these programs have been reported, one systematic review showed that P4P may be associated with improved process of care alone,3 whereas P4P has only limited and mixed effects on improvements in patients’ health outcomes.3,4 Other schemes, in contrast, may focus on the engagement of patients and simply reward participation in care-improvement activities, without necessarily linking bonuses to the attainment of objectives based on specific measures. This patient-focused approach (called ‘participatory’ throughout this article) places fewer limitations on professional autonomy and fosters cooperation between providers and other medical staff (e.g. health educators). Compared with previous studies on provider-focused P4P, studies on participatory/patient-focused P4P are less common and are at an early stage of development.5 As a result, it is worth investigating whether patient-focused P4P can affect quality of care.
The National Health Insurance Administration (NHIA) first announced the Enhanced Chronic Hepatitis B or C (Antiviral Therapy) Treatment Plan (abbreviation: drug plan) in 2003.6,7 Physicians voluntarily participating in the drug plan receive payments for using new antiviral drugs to treat severe hepatitis B or C. The enrollment criteria and definition of severe patients are provided in Appendix 1. Following the implementation of the drug plan, a P4P program was also implemented for treating hepatitis patients. This disease-specific P4P program (and other disease-specific P4P programs implemented in Taiwan) is designed as a participatory program, which usually has mixed features of physician- and patient-oriented systems. On the physician side, the programs mainly aim to improve process-based services using guidelines for physicians. On the patient side, the program’s main aim is disease management, such as providing health education and follow-up care. Taking the most well-established participatory diabetes P4P program initiated since 2001 as an example,8 several studies have reported that the P4P design is effective for treating diabetes with respect to certain process-based services and outcomes in terms of the clinical procedures performed (e.g. exams).9,10
Regarding this hepatitis P4P program launched in Taiwan in 2010, one recent study has found that this program is associated with a modest improvement in process, i.e. improved adherence to guidelines for physicians.11 However, whether this type of P4P has a positive effect on patient outcomes has not been examined and is thus still unknown. Hence, in the present study, we evaluated whether the patient-focused hepatitis P4P program can reduce negative health outcomes for patients treated with antiviral drugs, such as subsequent occurrence of liver cirrhosis and hospital admission.
Methods
Background of the new patient-focused P4P program
The prevalence of hepatitis B and C in Taiwan is high relative to the overall prevalence worldwide, with rates of 15% to 20% and 2% to 4%, respectively.12 Among hepatitis B patients, 60% to 90% have a risk of their disease progressing to liver cirrhosis and cancer.12 In 2009, the National Health Research Institute (NHRI) and the Gastroenterological Society of Taiwan published two clinical guidelines for liver disease in Taiwan, similar to the guidelines of the American Association for the Study of Liver Diseases (AASLD),13 with both suggesting that high-risk hepatitis patients should undergo one abdominal ultrasound examination every 6 months. However, according to the statistics of the National Health Insurance Administration (NHIA), in 2008 only 13% of hepatitis patients underwent the suggested abdominal ultrasound examination and glutamic-oxaloacetic transaminase (GOT)/glutamate pyruvate transaminase (GPT) examinations.14
The hepatitis P4P is a voluntary (participatory) P4P program in Taiwan called the Reimbursement Improvement Themes for the NHIA’s hepatitis B and C patients. This plan was initiated in January 2010 and is open to physicians specializing in gastroenterology, family medicine, or pediatrics who provide treatment for any patients with hepatitis. In addition to encouraging physicians to improve the quality of their care via extra incentives, this P4P plan has two purposes: (1) construct a patient-centered care (PCC) model and (2) encourage case managers to track patients through regular visits (every 6 months) according to the suggestions of the guideline. However, there are no detailed statements or restrictive regulations in the P4P plan that describe how to achieve these aims. If patients are enrolled by the participating physicians, then the physicians are potentially eligible for four types of payments per capita (US$1 = NT$30): new enrollment payment (US$3); payments associated with subsequent examinations per half a year (US$3); and other fees related to suspected cancer cases, including fees for transfer of suspected cancer cases checked by abdominal ultrasonography (US$17) and for early confirmation of suspected cancer cases (US$33).
Definition of hepatitis patients
A flow chart detailing the subject selection is shown in figure 1. This study’s data are derived from the NHIA’s databases for 2009–13. Following the protocol of the Reimbursement Improvement Themes for the NHIA’s hepatitis B and C patients, to be eligible as a hepatitis B or C patient from 2009 to 2013, participants must meet the following criteria: (1) having received a primary diagnosis of hepatitis B or C at least once (ICD-9-CM: 070.30–070.33, 070.51, 070.54, 571.40–571.49, 571.5, V02.61, V02.62, 070.30–070.33, 070.51, 070.54) (n = 98 026); (2) not having any cancer and/or not having been in a hepatic coma from 2006 to 2009 (n = 2,574); (3) not having died in the hospital during this period (n = 3126) and (4) having had at least two visits to the same hospital within 6 months (n = 44 285).11,14
Definition of severe patients in P4P (experimental group) and non-P4P (control group) groups
Patients had to have been enrolled in the P4P program (P42xx codes) (n = 8508). Among all 44 285 patients, 8508 patients (19%) had enrolled in the P4P program. This enrollment rate is similar to that reported in previous research.11 However, because Lin et al. (2016) revealed that the duration of the enrollment in the P4P program is associated with patient outcomes,15 we thus defined our P4P patients as ones who must have participated in the program in consecutive years from 2010 to 2013. Those P4P patients defined as such (n = 8508) were not only identified by the order code P4201C (new enrollment) in the 2010 NHIA claim data since the program was first implemented, but they must have also participated in the program over four consecutive years (P42xx codes, representing any two-digit figures after the prefix P42) (n = 916). All patients enrolled in the P4P program also participated in the drug plan. In other words, all P4P patients had severe hepatitis. In order to construct a comparable control group, we specifically analysed those patients in both the P4P (n = 916) and non-P4P groups (n = 1377) who participated in the drug plan in 2010. Patients in both groups had a similar level of disease severity.
We further employed the caliper matching method (also known as the greedy algorithm) to match patients in the P4P group with those in the potential comparison group on a one-to-one basis for the year 2010.16,17 This study used the demographic characteristics of patients, including age, gender and Charlson-Deyo Comorbidity Index (CCI), to generate a propensity score for each patient using a logistic regression. Because we applied the one-to-one matching scheme of the propensity score matching (PSM) method, the final sample size was 639 (70% of 916) for both the P4P and non-P4P groups.
Outcome measures
We used two health outcomes for the patients in this study—namely, the incidence (new cases) of liver cirrhosis and the rate of hospital admission for liver cirrhosis. The effects of the P4P program on these patient health outcomes were examined by comparing the two cohorts (P4P and non-P4P) in terms of their progress in developing liver cirrhosis (ICD: 571.5 or 456.1 or 456.21) and their hospital admission after follow-up from the index date in the year 2010. Any patient tracked before the index date who had a diagnosis of liver cirrhosis was excluded in the subsequent regression analysis.
Statistical analysis
The patient-level covariates included for analysis were as follows: participation status (primary independent variable), age, gender, CCI, income, job (including retirement), retirement (dichotomous variable), urbanization of location of residence, hepatitis B or C duration since 2006 and hospital-level factors. These factors included the nature of the hospital (tertiary, regional, or district hospital or stand-alone clinic), hospital ownership (public vs. private), teaching-hospital status, and treatment volume per year (the number of hepatitis patients treated by a hospital in a year). We stratified all districts into seven urbanization categories according to the standard published by Taiwan’s National Health Research Institute (NHRI).18
We utilized Cox proportional-hazards regression models with patient- and hospital-level factors for the year 2010 using backward elimination. We also used a multilevel model for sensitivity analysis (Appendix 2). The analysis was performed with SAS version 9.4 (Statistical Analysis Systems, Inc., Cary, NC, USA).
Results
Table 1 shows the baseline characteristics of the subjects in the experimental and control groups for the pre- and post-matched samples. Before the PSM, the patients receiving antiviral therapy in the two groups differed with respect to some of the demographic characteristics—namely, being older, being female and having more comorbidities were associated with not being enrolled in the P4P program. However, after the PSM, there were no significant differences between the two matched groups in terms of any observable demographic characteristics.
. | Prematched samples . | Postmatched samples . | ||||
---|---|---|---|---|---|---|
. | Control group . | Experimental group . | . | Control group . | Experimental group . | . |
. | n (%) . | n (%) . | P . | n (%) . | n (%) . | P . |
Total n | 1377(100) | 916(100) | 639(100) | 639(100) | ||
Age, y, (SEa) | 53(0.36) | 51(0.44) | <0.001 | 51(0.50) | 51(0.50) | 1.00 |
Charlson score, y, (SEa) | 1.95(0.03) | 1.24(0.03) | <0.001 | 1.39(0.04) | 1.39(0.04) | 1.00 |
Gender | 0.03 | 1.00 | ||||
Female | 541(39) | 401(44) | 242(38) | 242(38) | ||
Male | 836(61) | 515(56) | 397(62) | 397(62) | ||
Income (NT$) | 0.45 | 0.94 | ||||
≥$40 001 | 258(19) | 162(18) | 147(23) | 140(22) | ||
$20 001–$40 000 | 240(17) | 142(16) | 295(46) | 306(48) | ||
$1–$20 000 | 598(43) | 425(46) | 96(16) | 95(15) | ||
No incomeb | 281(20) | 187(20) | 101(16) | 98(15) | ||
Low incomec | 0.60 | 0.88 | ||||
No | 1367(99) | 911(100) | 638(100) | 637(100) | ||
Yes | 10(1) | 5(1) | 1(0) | 2(0) | ||
Job | 0.32 | 0.95 | ||||
Public servant | 71(5) | 57(7) | 37(6) | 42(7) | ||
Employee | 666(48) | 473(52) | 339(53) | 340(53) | ||
Farmer | 188(14) | 120(13) | 81(13) | 88(14) | ||
Fisherman | 30(2) | 18(2) | 16(3) | 16(3) | ||
Veterans & relatives | 38(3) | 17(2) | 12(2) | 10(2) | ||
No job or part-time job | 384(28) | 231(25) | 154(24) | 143(22) |
. | Prematched samples . | Postmatched samples . | ||||
---|---|---|---|---|---|---|
. | Control group . | Experimental group . | . | Control group . | Experimental group . | . |
. | n (%) . | n (%) . | P . | n (%) . | n (%) . | P . |
Total n | 1377(100) | 916(100) | 639(100) | 639(100) | ||
Age, y, (SEa) | 53(0.36) | 51(0.44) | <0.001 | 51(0.50) | 51(0.50) | 1.00 |
Charlson score, y, (SEa) | 1.95(0.03) | 1.24(0.03) | <0.001 | 1.39(0.04) | 1.39(0.04) | 1.00 |
Gender | 0.03 | 1.00 | ||||
Female | 541(39) | 401(44) | 242(38) | 242(38) | ||
Male | 836(61) | 515(56) | 397(62) | 397(62) | ||
Income (NT$) | 0.45 | 0.94 | ||||
≥$40 001 | 258(19) | 162(18) | 147(23) | 140(22) | ||
$20 001–$40 000 | 240(17) | 142(16) | 295(46) | 306(48) | ||
$1–$20 000 | 598(43) | 425(46) | 96(16) | 95(15) | ||
No incomeb | 281(20) | 187(20) | 101(16) | 98(15) | ||
Low incomec | 0.60 | 0.88 | ||||
No | 1367(99) | 911(100) | 638(100) | 637(100) | ||
Yes | 10(1) | 5(1) | 1(0) | 2(0) | ||
Job | 0.32 | 0.95 | ||||
Public servant | 71(5) | 57(7) | 37(6) | 42(7) | ||
Employee | 666(48) | 473(52) | 339(53) | 340(53) | ||
Farmer | 188(14) | 120(13) | 81(13) | 88(14) | ||
Fisherman | 30(2) | 18(2) | 16(3) | 16(3) | ||
Veterans & relatives | 38(3) | 17(2) | 12(2) | 10(2) | ||
No job or part-time job | 384(28) | 231(25) | 154(24) | 143(22) |
SE: Standard error.
No income: rely on parents.
Low income: these low-income patients are free from paying any NHIA insurance premium.
. | Prematched samples . | Postmatched samples . | ||||
---|---|---|---|---|---|---|
. | Control group . | Experimental group . | . | Control group . | Experimental group . | . |
. | n (%) . | n (%) . | P . | n (%) . | n (%) . | P . |
Total n | 1377(100) | 916(100) | 639(100) | 639(100) | ||
Age, y, (SEa) | 53(0.36) | 51(0.44) | <0.001 | 51(0.50) | 51(0.50) | 1.00 |
Charlson score, y, (SEa) | 1.95(0.03) | 1.24(0.03) | <0.001 | 1.39(0.04) | 1.39(0.04) | 1.00 |
Gender | 0.03 | 1.00 | ||||
Female | 541(39) | 401(44) | 242(38) | 242(38) | ||
Male | 836(61) | 515(56) | 397(62) | 397(62) | ||
Income (NT$) | 0.45 | 0.94 | ||||
≥$40 001 | 258(19) | 162(18) | 147(23) | 140(22) | ||
$20 001–$40 000 | 240(17) | 142(16) | 295(46) | 306(48) | ||
$1–$20 000 | 598(43) | 425(46) | 96(16) | 95(15) | ||
No incomeb | 281(20) | 187(20) | 101(16) | 98(15) | ||
Low incomec | 0.60 | 0.88 | ||||
No | 1367(99) | 911(100) | 638(100) | 637(100) | ||
Yes | 10(1) | 5(1) | 1(0) | 2(0) | ||
Job | 0.32 | 0.95 | ||||
Public servant | 71(5) | 57(7) | 37(6) | 42(7) | ||
Employee | 666(48) | 473(52) | 339(53) | 340(53) | ||
Farmer | 188(14) | 120(13) | 81(13) | 88(14) | ||
Fisherman | 30(2) | 18(2) | 16(3) | 16(3) | ||
Veterans & relatives | 38(3) | 17(2) | 12(2) | 10(2) | ||
No job or part-time job | 384(28) | 231(25) | 154(24) | 143(22) |
. | Prematched samples . | Postmatched samples . | ||||
---|---|---|---|---|---|---|
. | Control group . | Experimental group . | . | Control group . | Experimental group . | . |
. | n (%) . | n (%) . | P . | n (%) . | n (%) . | P . |
Total n | 1377(100) | 916(100) | 639(100) | 639(100) | ||
Age, y, (SEa) | 53(0.36) | 51(0.44) | <0.001 | 51(0.50) | 51(0.50) | 1.00 |
Charlson score, y, (SEa) | 1.95(0.03) | 1.24(0.03) | <0.001 | 1.39(0.04) | 1.39(0.04) | 1.00 |
Gender | 0.03 | 1.00 | ||||
Female | 541(39) | 401(44) | 242(38) | 242(38) | ||
Male | 836(61) | 515(56) | 397(62) | 397(62) | ||
Income (NT$) | 0.45 | 0.94 | ||||
≥$40 001 | 258(19) | 162(18) | 147(23) | 140(22) | ||
$20 001–$40 000 | 240(17) | 142(16) | 295(46) | 306(48) | ||
$1–$20 000 | 598(43) | 425(46) | 96(16) | 95(15) | ||
No incomeb | 281(20) | 187(20) | 101(16) | 98(15) | ||
Low incomec | 0.60 | 0.88 | ||||
No | 1367(99) | 911(100) | 638(100) | 637(100) | ||
Yes | 10(1) | 5(1) | 1(0) | 2(0) | ||
Job | 0.32 | 0.95 | ||||
Public servant | 71(5) | 57(7) | 37(6) | 42(7) | ||
Employee | 666(48) | 473(52) | 339(53) | 340(53) | ||
Farmer | 188(14) | 120(13) | 81(13) | 88(14) | ||
Fisherman | 30(2) | 18(2) | 16(3) | 16(3) | ||
Veterans & relatives | 38(3) | 17(2) | 12(2) | 10(2) | ||
No job or part-time job | 384(28) | 231(25) | 154(24) | 143(22) |
SE: Standard error.
No income: rely on parents.
Low income: these low-income patients are free from paying any NHIA insurance premium.
In table 2, the incidence densities of liver cirrhosis are 13‰ and 14‰ for the P4P and non-P4P groups, respectively, and the incidence densities of hospital admission are 4‰ and 9‰ for the P4P and non-P4P groups, respectively. The incidence densities of both of these negative outcomes over 4 years were thus lower in the P4P group. However, these results do not account for other confounders in the regression models.
. | Liver cirrhosis (n = 1116) . | Hospital admission (n = 1278) . | ||
---|---|---|---|---|
. | P4P . | Non-P4P . | P4P . | Non-P4P . |
Events | 22 | 25 | 8 | 18 |
Cases | 564 | 552 | 639 | 639 |
Person-years | 1748 | 1725 | 2004 | 2003 |
Incidence density (‰) | 13 | 14 | 4 | 9 |
. | Liver cirrhosis (n = 1116) . | Hospital admission (n = 1278) . | ||
---|---|---|---|---|
. | P4P . | Non-P4P . | P4P . | Non-P4P . |
Events | 22 | 25 | 8 | 18 |
Cases | 564 | 552 | 639 | 639 |
Person-years | 1748 | 1725 | 2004 | 2003 |
Incidence density (‰) | 13 | 14 | 4 | 9 |
. | Liver cirrhosis (n = 1116) . | Hospital admission (n = 1278) . | ||
---|---|---|---|---|
. | P4P . | Non-P4P . | P4P . | Non-P4P . |
Events | 22 | 25 | 8 | 18 |
Cases | 564 | 552 | 639 | 639 |
Person-years | 1748 | 1725 | 2004 | 2003 |
Incidence density (‰) | 13 | 14 | 4 | 9 |
. | Liver cirrhosis (n = 1116) . | Hospital admission (n = 1278) . | ||
---|---|---|---|---|
. | P4P . | Non-P4P . | P4P . | Non-P4P . |
Events | 22 | 25 | 8 | 18 |
Cases | 564 | 552 | 639 | 639 |
Person-years | 1748 | 1725 | 2004 | 2003 |
Incidence density (‰) | 13 | 14 | 4 | 9 |
In table 3, the Cox proportional-hazards regression models show that the P4P group had a lower risk of developing liver cirrhosis and being admitted to the hospital. However, the reduction in hospital admission (HR = 0.44, P = 0.05) and the reduction in cirrhosis (HR = 0.92, P = 0.77) were not statistically significant. The multilevel model also demonstrated similar results (Appendix 2). These findings indicate that the participatory P4P systems could not improve patient health outcomes in terms of reducing hospital admission and new cases of cirrhosis. In addition, the significant control variables in the two models were as follows: patient age (HR = 1.06, P < 0.001) in the cirrhosis model and the Charlson score (HR = 1.73, P = 0.005) in the admission model.
. | Cirrhosis model HR (95% CI) . | P . | Admission model HR (95% CI) . | P . |
---|---|---|---|---|
Participation | 0.92 (0.52,1.63) | 0.774 | 0.44 (0.19,1.01) | 0.052 |
Age | 1.06 (1.03,1.09)*** | <0.0001 | ||
Charlson score | 1.73 (1.18,2.54)** | 0.005 | ||
. | Cirrhosis model HR (95% CI) . | P . | Admission model HR (95% CI) . | P . |
---|---|---|---|---|
Participation | 0.92 (0.52,1.63) | 0.774 | 0.44 (0.19,1.01) | 0.052 |
Age | 1.06 (1.03,1.09)*** | <0.0001 | ||
Charlson score | 1.73 (1.18,2.54)** | 0.005 | ||
P < 0.01;
P < 0.001. CI: Confidence interval. HR: Hazard ratio.
An empty cell means that the variable was not significant for this model.
. | Cirrhosis model HR (95% CI) . | P . | Admission model HR (95% CI) . | P . |
---|---|---|---|---|
Participation | 0.92 (0.52,1.63) | 0.774 | 0.44 (0.19,1.01) | 0.052 |
Age | 1.06 (1.03,1.09)*** | <0.0001 | ||
Charlson score | 1.73 (1.18,2.54)** | 0.005 | ||
. | Cirrhosis model HR (95% CI) . | P . | Admission model HR (95% CI) . | P . |
---|---|---|---|---|
Participation | 0.92 (0.52,1.63) | 0.774 | 0.44 (0.19,1.01) | 0.052 |
Age | 1.06 (1.03,1.09)*** | <0.0001 | ||
Charlson score | 1.73 (1.18,2.54)** | 0.005 | ||
P < 0.01;
P < 0.001. CI: Confidence interval. HR: Hazard ratio.
An empty cell means that the variable was not significant for this model.
Discussions
According to our findings, the current form of the participatory P4P program implemented in Taiwan for severe hepatitis B and C patients is not associated with better health outcomes in terms of less frequent hospital admission and incidence of cirrhosis.
Because hepatitis B/C is a chronic ambulatory care sensitivity condition (ACSC), the UK’s Outcome Framework (UK’s P4P) has included ACSC for hepatitis as one of their indicators,19 and the empirical study of Harrison et al. also indicated that the introduction of P4P was associated with a decrease in ACSC.20 Our study yielded a different result indicating that P4P cannot reduce hospital admissions.
We did not observed a sufficient improvement in terms of reduced incidences of cirrhosis and hospital admission because of the P4P program in this study. One reason could be that the observation period was not long enough (4 years) with respect to the prognosis of the condition. Capturing this information depends on the availability of the data and could be further investigated when a longer period of data is available. Another reason could be the lack of a patient personalized care plan in the P4P program.
Including a personalized care plan in the P4P program, such as a disease management program (DMP), could be a good strategy.21 The concept of the DMP is derived from the chronic care model (CCM),22 which aims to transfer patients with chronic care from a regular clinical or reactive intervention to a planned and proactive initiative and to enhance patient skills related to self-management and activation. This model involves several activities,22 such as self-management support, promotion of clinical care that is consistent with concurrent scientific knowledge, visits that are planned to meet patient needs, regular follow-up, and full utilization of non-physician staff.
One example is the program implemented by the Italian region of Emilia-Romagna, which requires physicians to participate in the Diabetes Disease Program to be eligible for a bonus in addition to the original capitation payment; this program requires that certain activities be executed, including adherence to guidelines; performance of regular follow-up regarding lipids, cholesterol, HbA1c, blood pressure and medication dosage; dietary education; and foot examination.5 Relevant studies in Italy have demonstrated that this type of design can improve patient outcomes.5,17,23,24
As the leader of the health care team, the physician plays a crucial role in influencing patients’ behavior and thus enhancing patient outcomes. Because a physician-focused, clinical-process-oriented P4P program might overlook the management of the patient, other complementary/supplementary policies, such as a mandated DMP, could be implemented at the same time.
Because the hepatitis P4P program in Taiwan does not require a DMP to be set up by hospitals and only focuses on the physician side, PCC is currently lacking in the program (i.e. only the term ‘PCC’ is mentioned in the program without actually being implemented). In addition, even if a hospital wishes to hire a dedicated case manager to help medical teams execute DMP-required activities, the extra incentive is too small to hire one.25
There are some limitations to our study. First, PSM can only perform matching based on observable characteristics. Because not all important patient-related data were available in the NHIA’s databases, such as patient clinical test results (e.g. GOT or GPT values) or patient attitude, the possibility of selection bias in the experimental group cannot be completely eliminated. However, in this study, we did include important patient characteristics, such as age, gender, and comorbidity, in the propensity score matching and we believe that the selection bias has been kept to a minimum. In addition, although variables related to socioeconomic status, such as income, job, retirement, and urbanization of living, were not included in the propensity score analysis, the P4P and non-P4P groups were comparable because the two groups still showed similar distributions for these variables. Second, we do not have data on some of the objective outcomes, for example, mortality and subjective outcomes such as quality of life or patient satisfaction. However, the P4P program does not affect patient outcomes in terms of incidents of admission and cirrhosis and thus may have no effects on the corresponding outcomes, for example, mortality and quality of life. Third, the first version of the hepatitis P4P program proposal did not list the ICD codes for the primary diagnosis of hepatitis B or C, including V02.61 and V02.62, and until January 2011 these codes were suggested by the second version of the hepatitis P4P program proposal. To increase the P4P sample size in the present study, the hepatitis patients enrolled in the P4P program included those associated with these two ICD codes in 2010 (31 cases).
Conclusions
In summary, this study found that, in its current form, the participatory hepatitis P4P program has not resulted in reduced hospital admissions or new incidences of liver cirrhosis. The participatory P4P program could consider providing incentives and establishing requirements for including a disease management program to strengthen patient-centered care such that better patient health outcomes can be achieved in the future.
Acknowledgements
This study received ethics approval from the Institutional Review Board, Fu Jen Catholic University (IRB number C103158).
Funding
This work was supported by the Ministry of Science and Technology [NSC: 103-2628-H-030-003-] and Cathay General Hospital Medical Center [103‐CGH‐FJU‐18] in Taiwan.
Conflicts of interest: None declared.
P4P may be associated with improved process of care alone, whereas P4P has only limited and mixed effects in improving patients’ health outcomes.
Whether this participatory-focused hepatitis P4P program has a positive effect on the health outcomes of patients with severe hepatitis B or C has not been examined and is thus still unknown.
According to our findings, the current form of the participatory P4P program implemented in Taiwan for patients with severe hepatitis B or C is not associated with better health outcomes in terms of less frequent hospital admission and incidence of cirrhosis.
The participatory P4P program could consider providing incentives and requiring the inclusion of a disease management program to strengthen patient-centered care such that better patient health outcomes can be achieved in the future.
Comments