BIG DATA ANALYSIS ON INDONESIA’S PRIMARY CARE PAY-FOR PERFORMANCE

Pay-for-performance (P4P) schemes are commonly used to incentivize primary care providers, which is intended to improve the quality of healthcare they deliver. Challenges arise on how the scheme brings improvement (Zhang, Li, Yuan, & Zhu, 2024). P4P schemes provide financial incentives or facilities to health workers based on the achievement of predetermined performance goals. Various P4P programs have been implemented around the world. There is a question of which model is suitable for P4P implementation to achieve better results(Jamili et al., 2023). Several P4P programs in several countries had a limited impact on quality healthcare access rates and did not address health inequity (Katz et al., 2015). In several cases, P4P is good at driving some kinds of improvement but not others. In some areas, it also generated moral controversy, which in turn, created conflicts of interest for providers (Khan, Rudoler, McDiarmid, & Peckham, 2020). P4P scheme for provider payment is an innovative financing mechanism that may be similar to other type of payment systems such as results based financing, performance-based financing, performance-based contracting, output-based aid, conditional cash transfer and cash on delivery (Bauhoff et al., 2013; Musgrove, 2011). Incentives in the P4P scheme are provided to achieve a pre-agreed set of results (outputs and outcomes) or indicator targets. This is achieved by improving the performance of health workforce and health facilities. It also utilizes data monitoring in a stipulated time-frame . Traditionally, the achievement target in a P4P program is measured through health outcomes, utilization of services, and quality of care (Eichler, 2006; Soeters, 2006). Indonesia's Social Security Administrative Body for Health (BPJS Kesehatan) has applied Performance-Based Capitation (KBK) for primary care providers. Primary care providers are assessed every month before the due for capitation payment. Capitation payment will be adjusted ranging from 15 percent of total capitation received by the primary care providers. Previous indicators did not show fairness and only focus on process aspects. In 2019, an improvement in the KBK indicators was enacted. The indicators focused more on outcome impacts. The indicators were contact rate, non-specialty referral ratio, and proportion of-PROLANIS disease management program members with controlled clinical outcomes. With more than 22 thousand primary care providers having contract with Indonesia's Social Security Administering Body for Health Sector (BPJS Kesehatan) throughout the Indonesian archipelago, varieties of primary care providers in regards of available doctors and resourced facilities are inevitable. KBK is implemented to primary care providers with more than 5.000 registered JKN members with a minimum of a year of contract. To analyze the large amount of contracted healthcare facilities, the use of big data analysis is a must. Big data analysis techniques were introduced in 1997 and were identified by the “3Vs”: increasing volume of data, high velocity of data, and variety of data. In the following we use the term big data analysis as referring to the computational analysis of large data sets to assess patterns, trends, and associations in data collected from a wide range of sources in contrast to using classical statistics programs (Garapati et al., 2018; McAfee, 2012; Langkafel, 2014). In the KBK information system, numerous data variables will allow big data analysis on how the policy is implemented and the impacts on healthcare outcomes. Previous studies have not assessed the use of big data analysis in evaluating Indonesia’s KBK pay-for performance policy. This study aims to evaluate the effectiveness of the policy in terms of how the indicators show improvement. II. RESEARCH METHODOLOGY Study Design This study is a non-experimental big-data analysis by observational descriptive method. Data Sources Every primary care providers contracted by BPJS to record each healthcare service they deliver. Quantitative data were obtained from BPJS Kesehatan national database generated from healthcare services recorded by 10.729 primary care providers through P-Care application program. Analysis was conducted through BPJS Kesehatan Business Intelligence application program. Data from the program was collected by the data management system. Data Analysis All data from the P-Care program was prospectively collected and analysed through BPJS Kesehatan Business Intelligence and SSBI data management program software. Data was limited to primary care providers eligible for KBK capitation adjustment. Data series were analysed from 2020 to 2023. Further data from January to June 2024 were also analysed. Data analysis included type of primary care providers, area of the primary care providers, and target indicator achievement. Analysis based on descriptive statistics on the KBK data were generated to identify how the indicators show improvement in target achievements.

Study Design This study is a non-experimental big-data analysis by observational descriptive method. Data Sources Every primary care providers contracted by BPJS to record each healthcare service they deliver. Quantitative data were obtained from BPJS Kesehatan national database generated from healthcare services recorded by 10.729 primary care providers through P-Care application program. Analysis was conducted through BPJS Kesehatan Business Intelligence application program. Data from the program was collected by the data management system. Data Analysis All data from the P-Care program was prospectively collected and analysed through BPJS Kesehatan Business Intelligence and SSBI data management program software. Data was limited to primary care providers eligible for KBK capitation adjustment. Data series were analysed from 2020 to 2023. Further data from January to June 2024 were also analysed. Data analysis included type of primary care providers, area of the primary care providers, and target indicator achievement. Analysis based on descriptive statistics on the KBK data were generated to identify how the indicators show improvement in target achievements.

A. Healthcare Data Architecture in BPJS Kesehatan KBK Information System The business process automation in the KBK system relies on the data management which is crucial in determining the capitation received by primary care providers. First, the data is recorded in the P-Care application program. The data recorded is related on the healthcare service delivered by primary care providers including referral data. More than 1,6 million data each day is recorded through the P-Care application program. BPJS Kesehatan database will save the collected data. In order to make the data able to analyze, the Business Intelligence will do the calculations. KBK calculation is based on BPJS Kesehatan regulation number 7 year 2019. Each indicator has its own formula. BPJS Kesehatan data management system, through its Business Intelligence, analyzes the data from the PCare application program and concludes the performance of each provider every month. The data is processed in the BPJS Kesehatan data management system until the percentage amount of capitation adjustment is finally calculated before the due capitation payment. Several dashboards are also provided and can be monitored every month. Through the dashboard, BPJS Kesehatan can monitor each area and even each primary care provider. The dashboard can also monitor the adjusted capitation payment. Bi-weekly feedback is automated and can be accessed by primary care providers in the PCare application program. Primary care providers can login to their PCare account and monitor their achievement every two weeks so that they can immediately take actions to improve their performance. Capitation payment adjustment is automated based on the KBK Business Intelligence analysis. The percentage of adjustment ranges from 0–15% for Puskesmas and 0-5% for clinics. The calculation is done through automation on the KBK performance indicator achievement data. Another application program is BOA, which is used to pair the final KBK score to the capitation payment that the provider will receive. The complexity of the system resulted in numerous variables of data that were assessed in this study. Data were extracted from BPJS Kesehatan Business Intelligence and analyzed uniquely according to each indicator characteristic. B. Contact Rate Achievement The target for contact rate is ≥150‰. Higher contact rate shows more JKN members access to their providers. Lower rate shows the inability of primary care providers to deliver healthcare access to the JKN members. Until July 2024, three provinces have shown contact rate more than 150‰ (rating 4). These provinces are, DKI Jakarta(180,41‰), Bali (160,07‰), and West Sulawesi (154,89‰). Contact rate achievement from January to June 2024 is shown by Figure 1. FIGURE 1. PROVINCIAL DATA ANALYSIS ON CONTACT RATE TARGET INDICATOR ACHIEVEMENT Figure 1 also shows that provinces in the eastern part of Indonesia has lower achievement in contact rate, while provinces in the western part of Indonesia shows a higher achievement. Although provinces in the eastern part of Indonesia shows lower achievement, there has been improvement every year. Evaluation of contact rate in Papua, Maluku, and Nusa Tenggara provinces is shown by Figure 2. FIGURE 2. CONTACT RATE TREND IN PAPUA, MALUKU, AND NUSA TENGGARA PROVINCES FROM JANUARY 2020 TO DECEMBER 2023 Figure 2 shows Papua, Maluku, and Nusa Tenggara provinces have improved achievement in contact rate from 64,50% in January 2020 to 76,16% in December 2023. In a national level analysis for all types of primary care providers, from January 2020 to July 2024 there has been an improvement trend of contact rate achievement. In January 2020, contact rate was 128,57‰ while in July 2024 the rate increased to 149,56% as shown in Figure 3. FIGURE 3. NATIONAL TREND OF CONTACT RATE TARGET INDICATOR ACHIEVEMENT FROM JANUARY 2020 TO DECEMBER 2023 Furthermore, data analysis was enacted to assess contact rate achievement of each type of primary care provider. In July 2024, contact rate for primary care clinics (Klinik Pratama) has reached 156,28‰, for government owned health care centers (Puskesmas) 148,43‰ and primary care hospitals (RS Kelas D Pratama) reached 95,43‰. Overall, all types of FKTP shows a national average increasing trend. Figure 4 describes trend of contact rate achievement for each type of primary care provider from January 2020 to July 2024. FIGURE 4. NATIONAL TREND OF CONTACT RATE TARGET INDICATOR ACHIEVEMENT IN PUSKESMAS, KLINIK PRATAMA, AND RS KELAS D PRATAMA FROM JANUARY 2020 TO DECEMBER 2023 C. Non-Specialty Referral Ratio The target for non-specialty referral ratio is ≤2%. Lower percentage shows lower numbers of unnecessary referrals. There are six provinces that have not achieved the non-specialty referral ratio target until July 2024. These provinces are provinces located in the central and eastern part of Indonesia. Contact rate data on these six provinces is shown by Table 1. TABLE 1. PROVINCES NOT REACHING NON-SPECIALTY REFERRAL RATIO TARGET FROM JANUARY TO JULY 2024 Provinces Contact Rate January To July 2024 Kalimantan Tengah 4,15% Maluku 3,38% Nusa Tenggara Timur 4,01% Papua 5,17% Papua Barat 5,30% Sulawesi Tengah 4,23% Overall achievement shows improvement every year from 1,87% in January 2020 to 0,85% in July 2024. This is shown by Figure 5. As seen in Figure 5, the national trend on the non-specialty referral ratio shows a more progressive improvement than that of the contact rate FIGURE 5. NATIONAL TREND ON NON-SPECIALTY REFERRAL RATIO TARGET INDICATOR ACHIEVEMENT FROM JANUARY 2020 TO JULY 2024 We conducted further analysis on referral ratio data through the BPJS Kesehatan SSBI application program. Analysis of referral data from 2020 to 2023 shows there has been an increase in the proportion of non-specialty referrals with time, age, comorbidity, and complications compared to total non-specialty referrals. In 2020, the proportion of non-specialty referrals with TACC criteria was 54,48%. The proportion increased to 71.1% in 2023. Data of non-specialty referrals with TACC criteria can be seen in Table 2. TABLE 2. PROPORTION OF NON-SPECIALTY REFERRALS WITH TACC CRITERIA COMPARED TO TOTAL NON-SPECIALTY REFERRALS Year Referrals With TACC Criteria Total NonSpecialty Referrals Proportion Of Non-Specialty Referrals With TACC Criteria 2020 675.035 1.480.067 54,4% 2021 394.361 1.163.155 66,1% 2022 392.136 1.261.923 68,9% 2023 408.176 1.410.359 71,1% Further analysis of referral data through the BPJS Kesehatan SSBI application program shows the nonspecialty referral ratio in Puskesmas was better than that of Klinik Pratama. From January 2020 to July 2024, the non-specialty referral ratio in Puskesmas was 0,91%, while in Klinik Pratama the number was 1,13% and RS Kelas D Pratama 6,07%. Trends per month from January 2020 to July 2024 are shown by Figure 6. FIGURE 6. NATIONAL TREND ON NON-SPECIALTY REFERRAL RATIO TARGET INDICATOR ACHIEVEMENT IN PUSKESMAS, KLINIK PRATAMA, AND RS KELAS D PRATAMA FROM JANUARY 2020 TO JULY 2024 D. Proportion of PROLANIS Disease Management Program Members with Controlled Clinical Outcome The target for proportion of PROLANIS disease management program members with controlled clinical outcome (PROLANIS indicator) is ≥5%. Higher achievements show better chronic disease management in the primary care providers. From January to July 2024, eight provinces have reached indicator targets. Three of these Provinces are from Java Island; D.I. Yogyakarta, DKI Jakarta, and Jawa Timur. However, the highest indicator target achievement was shown by Lampung, Sulawesi Barat, and Bali as shown by Table 3. TABLE 3. PROVINCES THAT HAS REACHED PROLANIS INDICATOR TARGET FROM JANUARY TO JULY 2024 No Province Prolanis Indicator Achievement 1 Lampung 7,56% 2 Sulawesi Barat 7,37% 3 Bali 5,37% 4 D.I. Yogyakarta 5,33% 5 DKI Jakarta 5,32% 6 Jawa Timur 5,26% 7 Sulawesi Selatan 5,10% 8 Sumatera Utara 5,09% Overall achievement shows improvement every year from 1,72% in January 2020 to 4,71% in July 2024. This is shown by Figure 7. Further analysis on the PROLANIS indicator for each type of primary care provider shows that klinik pratama has the highest score (4,13%), followed by Puskesmas (3,21%), and RS Kelas D Pratama (1,17%) as shown in Figure 8. FIGURE 7. NATIONAL TREND OF PROLANIS INDICATOR TARGET ACHIEVEMENT FROM JANUARY 2020 TO JULY 2024 FIGURE 8. NATIONAL TREND ON PROLANIS TARGET INDICATOR ACHIEVEMENT IN PUSKESMAS, KLINIK PRATAMA, AND RS KELAS D PRATAMA FROM JANUARY 2020 TO JULY 2024 KBK data not only shows indicator target achievement but also shows several data insights. Data on JKN members diagnosed with diabetes mellitus (DM) and hypertension (HT) were analyzed. The following table shows five municipalities with the highest proportion of JKN members diagnosed with DM. TABLE 4. MUNICIPALITIES WITH THE HIGHEST PROPORTION OF JKN MEMBERS DIAGNOSED WITH DM YEAR 2023 No Municipality Registered JKN Members JKN Members Diagnosed with DM Proportion of JKN Members Diagnosed with DM 1 Kota Mojokerto 154.634 8.527 5,51% 2 Kota Madiun 213.119 10.437 4,90% 3 Kab. Kep. Seribu 28.985 1.310 4,52% 4 Kota Tomohon 92.213 4.085 4,43% 5 Kota Yogyakarta 441.646 19.348 4,38% Of the five municipalities in Table 4, one is a municipality outside Java Island. This is Kota Tomohon, located on Sulawesi Island. The following table shows five municipalities with the highest proportion of JKN members diagnosed with hypertension. Table 5 shows that municipalities outside Java Island have the highest proportion of JKN members diagnosed with HT (Kota Tomohon and Kabupaten Minahasa). TABLE 5. MUNICIPALITIES WITH THE HIGHEST PROPORTION OF JKN MEMBERS DIAGNOSED WITH HYPERTENSION YEAR 2023 No Municipality Registered JKN Members JKN Members Diagnosed with HT Proportion of JKN Members Diagnosed with HT 1 Kota Tomohon 92.213 16.745 18,16% 2 Kab. Minahasa 222.832 32.124 14,42% 3 Kota Magelang 137.006 17.948 13,10% 4 Kota Yogyakarta 441.646 54.658 12,38% 5 Kab. Hulu Sungai Tengah 230.550 28.021 12,15% E. Healthcare Data Architecture in BPJS Kesehatan KBK Information System The business process automation in the KBK system relies on the data management, which is crucial in determining the capitation received by primary care providers. First, the data is recorded in the P-Care application program. The data recorded is related to the healthcare service delivered by primary care providers, including referral data. More than 1,6 million data each day are recorded through the PCare application program. BPJS Kesehatan database will save the collected data. In order to make the data able to be analyzed, the Business Intelligence will do the calculations. KBK calculation is based on BPJS Kesehatan regulation number 7 year 2019. Each indicator has its own formula. BPJS Kesehatan data management system, through its Business Intelligence, analyzes the data from the PCare application program and concludes the performance of each provider every month. The data is processed in BPJS Kesehatan data management system until the percentage amount of capitation adjustment is finally calculated before the due capitation payment. Several dashboards are also provided and can be monitored every month. Through the dashboard, BPJS Kesehatan can monitor each area and even each primary care provider. The dashboard can also monitor the adjusted capitation payment. Bi-weekly feedback is automated and can be accessed by primary care providers in the PCare application program. Primary care providers can login to their PCare account and monitor their achievement every two weeks so that they can immediately take actions to improve their performance. Capitation payment adjustment is automated based on the KBK Business Intelligence analysis. The percentage of adjustment ranges from 0–15% for Puskesmas and 0-5% for clinics. The calculation is done through automation on the KBK performance indicator achievement data. Another application program is BOA, which is used to pair the final KBK score to the capitation payment that the provider will receive. The complexity of the system resulted in numerous variables of data that were assessed in this study. Data were extracted from BPJS Kesehatan Business Intelligence and analyzed uniquely according to each indicator characteristic.

The Indonesian JKN program relies on PCPs as healthcare gatekeepers in a multitiered referral system. They are expected to deliver quality healthcare in a cost effective structure. The three KBK indicators were meant to support quality healthcare in the primary care settings. In order to evaluate the healthcare services performed by the PCPs, big data analysis were used in this study. This is inline with other studies that have shown the use of big data analysis to evaluate health care policies. A study by Moutselos et al (2020) has shown a public health policy model evaluation using big data analytics in Sarajevo, Bosnia, and Herzegovina (Moutselos & Maglogiannis, 2020). A similar approach was conducted in this study. Data on each indicator was evaluated with data variables unique to each indicator characteristic. Pay-for-performance is related to equity implications (Miller, G. et al, 2013). Contact rate is intended to delivery equity in healthcare accessibility and ensure primary care providers act as gatekeepers. In the KBK data, primary care providers show an improvement in contact rate. This means that more access has been achieved and more JKN members have benefited from the primary care healthcare. However, since registered members increase every year, the challenges are how to provide sufficient numbers of healthcare workers to contact patients and deliver quality healthcare. The contact rate of klinik pratama has reached the intended target and shows a better performance than that of Puskesmas. This may be due to the numbers of JKN members in klinik pratama is much less than that of Puskesmas. For RS Kelas D Pratama, contact rate shows the least improvement but this is due to the lack of geographical ease of access. Most RS Kelas D Pratama are located in remote areas that need improvement in infrastructure. Since payment for primary care is by capitation, while payment for hospital care is by a semi fee-forservice through case-based-group method, there is the risk of excess referrals and high expenditures for specialty care (Saltman, R.B. et al., 2006). Nonspecialty referral ratio is intended to control primary care unnecessary referrals. This study shows nonspecialty referral ratio for all types of primary care providers has shown improvement. However, total amount of referrals and the use of TACC criteria have increased. This increase was apparent since the enactment of KBK policy. More referrals using TACC criteria may mean that more JKN members show complications of their disease. However, this needs to be validated. The national medical guideline for primary care has not yet encompassed TACC criteria for all disease which should be treated in the primary care settings. For this case, audit on primary care referrals using TACC criteria is needed. Moreover, for primary care providers in rural areas or outside Java island, the government must provide more healthcare workers and medical supplies in order to strengthen primary care providers and to avoid unnecessary referrals. For the PROLANIS indicator, all types of primary care providers has shown improvement with the highest score achieved by klinik pratama. This may be due to the numbers of registered JKN members in klinik pratama are much less than Puskesmas, therefore allowing healthcare workers to focus better on healthcare monitoring. Challenges of this indicator vary from the awareness of PROLANIS members to control their lifestyle to the availability of healthcare workers and essential medical supplies to treat patients with DM and HT. Although several provinces outside Java show better PROLANIS achievement, several municipalities outside Java shows higher proportion of JKN members diagnoses with DM and HT. Cultural customs affecting diet and lifestyle may be a potential factor. Therefore there must be an intensive health promotion prioritizing in areas with high proportion of JKN members diagnoses with DM and HT. Through big data analysis, KBK indicator achievement vary from one region to another. Early assumptions indicated that providers in Java and Sumatra will show better performances due to better healthcare accessibility and better healthcare infrastructure. However, this is not always the case. Regions outside Java and Sumatra have also shown better KBK indicator achievement implying that there are factors other than accessibility that may become an enabler to achieve the indicator target. One major factor is the fulfillment of patient expectations by the general practitioners in PCPs. Several studies have shown that how patient expectations are fulfilled by good PHC although most studies were analyzed in a broad perspective and were dominantly studied in Europe and North America (Kleij et al., 2017; Jung et al., 2003; Grol, 1999). For example, a study in the Netherland concluded that patients GP delivering sufficient consultation time, schedule availability, and delivering detailed information on patients’ illness resulted in better satisfaction (Jung et al., 1997). PCPs outside of Java and Sumatra or in rural areas might have more time to communicate with patients and address the therapy more thoroughly. These regions have lesser population, therefore, general practitioners have more time to analyze patients health status and deliver treatment properly leading to better contact to patients, better referral selection and more effective patient monitoring. Nevertheless, most of the regions show less KBK indicator achievement. Factors related to geographical accessibility, health facilities availability, limited general practitioner, cultural customs, communication data coverage, and support from local governments were potential factors that may affect all indicator achievement. The JKN program is a multistakeholder program, hence the support from all stakeholders is essential. Since it is a descriptive-method study through big data analysis, this study has several limitations. Data retrieved may not reflect the real situation. Healthcare data was recorded by primary care provider administrators in the BPJS Kesehatan PCare application program. Under or over reported data may be an inherent issue. We were unable to justify the validity and quality of services by the primary care providers. Therefore, quality of the electronic data were prone to human errors. Future studies should examine by more in-depth analyses on the quality of health services conducted by primary care providers in the KBK scheme.

From the KBK electronic data, primary care providers performance can be analyzed. In addition, other related data insights may also be analyzed. Bigdata analysis on the KBK scheme showed that there has been improvement in the indicator target achievement. From the three indicators, only nonspecialty referral ratio has reached the intended target. Several factors may affect the achievement: the number of registered JKN members in each primary care provider, the number of healthcare workers delivering services, availability of medical supplies, geographical factors, and most importantly the support from both central and local authorities. It is essential that the KBK data is used to evaluate the policy further.