BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from

Methods and Rationale of a Multi-Centre Cross Sectional Costing Study of Elective Percutaneous Coronary Intervention: A Study Design based on , a Middle- income Country

ForJournal: peerBMJ Open review only

Manuscript ID bmjopen-2016-014307

Article Type: Research

Date Submitted by the Author: 16-Sep-2016

Complete List of Authors: Lee, Kun Yun; University of Malaya, Department of Social and Preventive Medicine Ong, Tiong Kiam; Heart Centre, Division of Cardiology Low, Ee Vien; Bahagian Perkhidmatan Farmasi Kementerian Kesihatan Malaysia Liau, Siow Yen; Queen Elizabeth 2 Hospital, Department of Pharmacy Anchah, Lawrence; Sarawak General Hospital Heart Centre, Department of Pharmacy Hamzah, Syuhada; General Hospital , Administrative Office Mohd Ali, Rosli; National Heart Institute Liew, Huong Bang; Queen Elizabeth 2 Hospital, Division of Cardiology http://bmjopen.bmj.com/ Ismail, Omar; Penang General Hospital , Division of Cardiology Wan Ahmad, Wan Azman; University Malaya Medical Centre, Medicine Said, Mas Ayu; University of Malaya, Department of Social and Preventive Medicine Dahlui, Maznah; University of Malaya, Department of Social and Preventive Medicine

Primary Subject Health economics Heading: on October 1, 2021 by guest. Protected copyright. Secondary Subject Heading: Cardiovascular medicine, Health services research, Public health

Coronary heart disease < CARDIOLOGY, Coronary intervention < Keywords: CARDIOLOGY, STATISTICS & RESEARCH METHODS, Low- and middle- income countries, Costing approach

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 1 of 22 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 TITLE 4 5 Methods and Rationale of a Multi-Centre Cross Sectional Costing Study of Elective Percutaneous 6 7 Coronary Intervention: A Study Design based on Malaysia, a Middle-income Country. 8 9 10 AUTHORS 11 12 Lee Kun Yun a *, Ong Tiong Kiam b, Low Ee Vien c, Liau Siow Yen d, Lawrence Anchah e, Syuhada 13 f g h i j 14 Hamzah , Liew Huong Bang , Rosli Mohd Ali , Omar Ismail , Wan Azman Wan Ahmad , Mas 15 Ayu Said a, Maznah Dahlui a 16 For peer review only 17 18 a Department of Social and Preventive Medicine, Faculty of Medicine, University of Malaya, 19 Malaysia. 20 21 b Division of Cardiology,, Sarawak General Hospital Heart Centre, Malaysia. 22 23 c 24 Pharmaceutical Services Division, Ministry of Health, Malaysia. 25 d 26 Department of Pharmacy, Queen Elizabeth 2 Hospital, Sabah, Malaysia. 27 28 e Department of Pharmacy, Sarawak General Hospital Heart Centre, Malaysia. 29 30 f Administrative Office, Penang General Hospital, Malaysia. 31 32 g Division of Cardiology, Queen Elizabeth 2 Hospital, Sabah, Malaysia. 33 http://bmjopen.bmj.com/ 34 h National Heart Institute, Kuala Lumpur, Malaysia 35 36 i 37 Division of Cardiology, Penang General Hospital, Penang, Malaysia. 38 j 39 Department of Medicine, Faculty of Medicine, Kuala Lumpur, Malaysia. 40 41 *Corresponding Author on October 1, 2021 by guest. Protected copyright. 42 43 44 Email address: [email protected] Telephone number: + 603-7967 4756 45 46 47 WORD COUNT 3975 48 49 50 KEYWORDS 51 52 Costing approach, research methods, coronary intervention, low- and middle-income countries 53 54 55 56 57 58 59 1 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 2 of 22

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 ABSTRACT 4 5

6 7 Objectives: 8 9 Limitations in the quality and access of cost data from low- and middle-income countries constrain 10 11 the implementation of economic evaluations. With the increasing prevalence of coronary artery 12 disease in Malaysia, local cost information are vital for cardiac service expansion. We aim to 13 14 describe a modified costing data collection method tailored to our country setting, and to present the 15 preliminary result. Percutaneous coronary intervention (PCI) was chosen for this purpose. 16 For peer review only 17 18 Design: 19 20 This is a cross-sectional costing study from the perspective of healthcare providers, using top-down 21 approach, from January-June 2014. Cost items under each unit of analysis involved in the provision 22 23 of PCI service were identified, valuated and calculated to produce unit cost estimates. 24 25 26 Setting: 27 28 Five public cardiac centres participated. All the centres provide full-fledged cardiology services. 29 30 They are also the tertiary referral centres of their respective regions. 31 32 Participants: 33 http://bmjopen.bmj.com/ 34 35 We included only patients who undergo PCI as an elective procedure. Patients with urgent/emergent 36 indication for PCI, or with shock and hemodynamic instability were excluded. 37 38 39 Primary and secondary outcome measures: 40

41 on October 1, 2021 by guest. Protected copyright. 42 The outcome measures of interest were the unit costs of admission to cardiac ward and cardiac 43 catheterization utilization, which made up the PCI hospitalization cost. 44 45 46 Results: 47 48 Preliminary results from a single centre showed the total hospitalization cost per patient for elective 49 50 PCI to be RM 14,309.01 (USD 3974.73). PCI consumables were the dominant cost item. 51 52 53 54 55 56 57 58 59 2 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 3 of 22 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 4 Conclusions: 5 6 Comprehensive results from all centres enable comparison from the levels of cost items, unit costs to 7 8 total costs. Standardized costing approach in a multi-centre study generates important information on 9 10 cost variations between centres, thus providing valuable guidance for service planning. In the setting 11 of limited data availability like many LMIC, a modified costing method can be used for the purpose 12 13 of economic evaluations. 14 15 16 Registration: For peer review only 17 18 Malaysian MOH Medical Research and Ethics Committee (ID: NMRR-13-1403-18234 IIR) 19 20 21 22 ARTICLE SUMMARY 23 24 Strengths and limitations of this study 25 26 27 • Based on limitations in local settings, modified costing approach can be applied in LMIC to 28 29 enhance the uptake of economic evaluation for healthcare services. 30 • Cost estimates from standardized collection methods can lead to between-centre comparison 31 32 at multiple levels from individual cost items to overall hospitalization cost.

33 http://bmjopen.bmj.com/ 34 • The non-participation of private cardiac centres may limit the generalizability of the results. 35 36 37 INTRODUCTION 38 39 40 Economic evaluation is an important component of modern healthcare delivery. With the increasing on October 1, 2021 by guest. Protected copyright. 41 prevalence of diseases and advancement of medical technologies, healthcare costs continue to 42 43 escalate. In many developed countries, there are well-established guidelines for healthcare economic 44 45 evaluation.[1] Most of the guidelines available were established in the backdrop of an easy access to 46 high quality and freely available data, based on the setting in high-income countries. Therefore, 47 48 costing analysis of healthcare services are commonly applied to assess the impact of investments in 49 50 disease prevention and treatment, and to guide decision-making in future budget allocation and 51 service planning.[2] In contrast, the lack of infrastructure and financial support for evidentiary data 52 53 capture in low and middle-income countries (LMIC) results in a paucity of easily accessible and 54 reliable cost data. This becomes a huge challenge to the healthcare providers in these countries 55 56 towards the uptake of research and application of economic evaluation in their daily practice.[3-6] 57 58 59 3 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 4 of 22

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 4 As with many other LMIC, the total health expenditure in Malaysia has been steadily increasing 5 6 from 3.0% in 2000, 4.0% in 2013 to 4.2% of national gross domestic product in 2014[7]. Public and 7 private healthcare sector expenditures were almost equal at 52.3% and 47.4%.[7, 8] However, the 8 9 expenditure pattern is not a good reflection of the difference in utilization rate between the public 10 and private sectors. Public healthcare service in Malaysia recorded higher utilization because it is 11 12 provided free or at a highly subsidized rate. This was reflected in the number of admission that was 13 14 twice as higher to public hospitals compared to private hospitals.[9] 15 16 For peer review only 17 Among the various disease burdens in Malaysia, cardiovascular disease (CVD) is a leading cause of 18 19 morbidity and mortality. While the CVD mortality rate is declining in developed nations, the drastic 20 21 urbanization and lifestyle changes brought on by epidemiological transition has led to higher 22 prevalence of non-communicable diseases in LMIC, with CVD topping the list.[10] As a chronic 23 24 condition requiring frequent follow-ups and hospitalization, CVD places a huge economic burden on 25 the healthcare system. Hospitalizations and cost burden attributed to CVD have shown an increasing 26 27 trend in the European countries.[11-13] Malaysia is not spared of the same CVD epidemic. In 2013, 28 29 CVD accounted to 24.71% of total mortality in Malaysia and one of the top five causes of 30 hospitalization.[9] 31 32

33 http://bmjopen.bmj.com/ 34 CVD encompasses a range of medical conditions such as coronary artery disease (CAD), 35 cerebrovascular disease, peripheral artery disease, and heart failure. Among all, CAD accounts for 36 37 the highest prevalence and mortality. CAD can present as mild angina attack to potentially fatal 38 acute coronary syndrome, a spectrum ranging from unstable angina to myocardial infarction. 39 40 Percutaneous coronary intervention (PCI) is a common treatment modality for CAD due to its safety 41 on October 1, 2021 by guest. Protected copyright. 42 profile in terms of lower mortality and complications.[14, 15] In many countries, the availability of 43 coronary catheterization facilities and access to PCI is a key performance indicator of the cardiology 44 45 service. However, the need for sophisticated laboratory facilities, highly skilled clinical staff and 46 47 costly consumables such as cardiac stents often drive up the cost of PCI. This can be a constraining 48 factor towards its service establishment and delivery in resource-limited countries. 49 50 51 52 With the increasing incidence of CAD, expansion of PCI services across Malaysia is underway. 53 More than 60 public, private and teaching institutions are performing approximately 12000 PCIs 54 55 annually in Malaysia, majority being elective PCI. In the public healthcare system, cardiology 56 57 58 59 4 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 5 of 22 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 services are mainly provided by Ministry of Health teaching hospitals and state-level hospitals. In 4 addition, Malaysia has a highly specialized cardiac centre, which was established as a corporatized 5 6 entity, and given autonomy in terms of financial an staff autonomy, even though it is fully owned by 7 8 the government. 9 10 11 Cost of PCI provision will continue to escalate with the advancement in medical technologies, as 12 13 with many other cardiac services. The scarcity of health resources necessitates cost-saving 14 mechanisms to be put in place. Policy makers and healthcare professionals will be interested to gain 15 16 more insight into theFor cost variation peer of performing review cardiac procedures only such as PCI in different 17 centres. In order to do that, reliable cost data from different cardiac centres are needed. However, 18 19 local evidence especially among public hospitals are lacking; as the development of health 20 21 economics is still in its infancy stage in Malaysia and efforts to conduct economic evaluations are 22 hampered by limited access to financial data and non-computerized patient clinical data. 23 24 25 26 OBJECTIVES 27 28 29 With these in mind, we describe a simple, standardized data collection method for a multi-centre 30 costing study of cardiac service in Malaysia. Elective PCI is the chosen procedure as it is among the 31 32 most commonly conducted procedure with the least heterogeneity among the patients and providers. 33 http://bmjopen.bmj.com/ 34 We aim to outline the modifications made to existing guidelines that were developed and 35 implemented in HIC. By presenting the preliminary result from one of the participating centre, we 36 37 hope to illustrate that it is feasible to conduct a comprehensive costing analysis in a LMIC setting 38 39 using modified costing approach. 40 41 on October 1, 2021 by guest. Protected copyright. 42 METHODS 43 44 45 Study Design 46 47 48 This costing analysis was designed as a hospital-based cross-sectional study, from the perspective of 49 st th 50 healthcare providers. Cost data collection was conducted from January 1 2014 to June 30 2014; 51 using a top-down costing approach. All cost estimates were presented in the local currency, 52 53 Malaysian Ringgit (RM), whereby USD 1=RM 3.60 at the time of study. 54 55 56 57 58 59 5 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 6 of 22

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 4 Patient Population and Clinical Data Collection 5 6 7 8 PCI can be conducted as an emergency or elective procedure. Patients who undergo emergency PCI 9 often have more severe comorbidities and disease presentation, leading to higher resource 10 11 consumption and hospitalization cost. They often develop complications post PCI thus requiring 12 13 admission to an intensive care unit. On the contrary, patients who undergo elective PCI have lower 14 risk profile and likely have fewer complications. As a relatively homogenous group, patients who 15 16 had elective PCI wouldFor be a better peer proxy to analyzereview hospital-level only cost. As such, we restricted our 17 analysis to patients who undergone PCI as an elective procedure, and excluding all patients with 18 19 urgent/emergent indication for PCI, or with shock and hemodynamic instability. This definition is 20 21 compatible with the definition of elective PCI adopted by other studies in the literature.[16, 17] 22 23 24 Study Sites 25 26 27 For the cost data collection, all the public and private cardiac centres that provided PCI service were 28 invited to collaborate in this study. However, only five public cardiac centres agreed to participate. 29 30 All five are the tertiary referral hospitals of their respective regions and they provide full-fledged 31 cardiology and cardiothoracic services. Each was staffed by at least one interventional cardiologist. 32 33 Three of the study sites (Centre II, III, and IV) are government public hospitals whereas Centre I is a http://bmjopen.bmj.com/ 34 35 semi-corporatized university teaching hospital. Centre V is a specialized heart centre established as a 36 corporatized entity. The range of centres was able to represent variations across the different cost 37 38 components stemming from the heterogeneity in hospital characteristics, treatment preferences and 39 40 geographical locations. Table 1 outlines the main description of each centre. 41 on October 1, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 6 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 7 of 22 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 Table 1 Background Information of Study Centres. 4 5 6 Centre I II III IV V 7 8 Ownership Type Semi- Public Public Public Corporatized 9 corporatized 10 Start of PCI operation 1987 1994 2011 2010 1992 11 12 13 Interventional Cardiologist 5 3 1 1 15 14 15 Fellow Cardiologists 12 8 7 0 12 16 For peer review only 17 *Total Hospital Admission 34414 29336 1744 9340 8485 18 19 20 *Total Procedures in Cardiac 1084 1118 1485 940 2709 21 Catheterization Laboratory 22 *Total PCI Procedures 580 257 282 251 1558 23 24 *During the study period of January-June 2014 25 26 27 28 29 30 31 32

33 http://bmjopen.bmj.com/ 34 35 36 37 38 39 40

41 on October 1, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 7 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 8 of 22

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 Cost Data Collection 4

5 6 Due to the inherent characteristics of financial and medical record keeping in each healthcare system, 7 8 there is no pre-existing data collection tool suitable for use by all. International guidelines 9 recommended the design of a standardized cost collection tool that should be universally accessible 10 11 and user-friendly for all involved parties.[18, 19] Thus, the health economists and interventional 12 13 cardiologists in our study created a data collection tool using MS Excel. Improvements were made 14 based on the feedbacks from the pilot study at one of the centres. A training workshop was organized 15 16 for the data enumeratorsFor and studypeer coordinators, review covering topics only such as the purpose of the study, 17 introduction of the data collection tool, definitions of each of the cost categories, instructions on how 18 19 to use the tool, and troubleshooting. Cost data collected were assessed for consistency and validity. 20 21 Any queries on incomplete data were clarified with the respective centre. 22 23 24 Costing Pathway (Figure 1) 25 26 27 Step 1. Identification of unit of analysis 28 29 To achieve the most accurate estimate of PCI cost, it is important to identify all the resources 30 31 consumed by the patients during hospital admission. Many HIC used Diagnosis Related Group 32

33 (DRG), a system which classifies acute inpatient episodes based on the main clinical condition, as http://bmjopen.bmj.com/ 34 unit of analysis for illness-specific costing.[20] However, DRG is not practiced in any of the study 35 36 centres. As an alternative, we outlined the event pathway for the provision of care for PCI patients 37 from admission to discharge. Two main units of analysis were identified, namely cardiac ward (CW) 38 39 and cardiac catheterization laboratory (CL). All the cost items under each unit of analysis were 40

41 categorized under direct medical cost or overhead cost. on October 1, 2021 by guest. Protected copyright. 42 43 44 Step 2. Identification of Cost Items 45 46 47 For CW, the four cost items included under direct medical costs were labor, capital, consumables 48 49 and medications; whereas overhead costs included utility, dietary, ancillary service and hospital 50 support service. There were less cost items under CL, the second unit of analysis. Medications were 51 52 prescribed and served at CW, thus not taken as a cost item under CL. Dietary and ancillary support 53 service were also provided only once under CW. 54 55 56 57 58 59 8 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 9 of 22 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 Step 3. Valuation of Cost Items 4 5 6 In HIC, reference cost from national database or pre-set cost-to-charge ratios are used for cost item 7 valuation in economic evaluation, such as the NHS reference cost in the United Kingdom,[21] and 8 9 standard cost in the Australian costing guide.[22] However, reference cost is still under development 10 in Malaysia. Thus, actual calculation of unit cost for all the cost items was conducted using a top- 11 12 down approach. The subsequent sections discussed the valuation of each cost item. 13 14 15 Labour 16 For peer review only 17 This referred to the sum of full salary inclusive of wages and employers social contributions paid for 18 19 the staff. A list of all the clinical staff and their respective grades involved in the provision of PCI 20 service were used to obtain their respective payroll register from the hospital finance department. 21 22 The average salary of each position was calculated from the maximum and minimum monthly 23 salaries. 24 25 26 27 For labour cost of clinical staff stationed full time in the same unit, their salaries were summed up in 28 total as all of their productivity were assumed to be contributed to the units they were stationed in. 29 30 This included nurses and attendants in CW, laboratory technicians and radiographers in CL. 31

32 33 As for doctors, their work scope may spread across several units; including conducting interventions http://bmjopen.bmj.com/ 34 35 in the CL, treating patients in the specialist clinic, and leading the clinical ward rounds. Therefore, 36 expert opinion from the interventional cardiologists at each centre was sought on the portion of time 37 38 spent on different activities by doctors at each cardiac centre. The apportioned labour cost of doctors 39 40 based on working hours assumed that all doctors were equally productive in delivering patient- 41 related care and no idle time was spent. on October 1, 2021 by guest. Protected copyright. 42 43 44 Capital 45 46 This included fixed, one-time expenses incurred on the purchase of land, building, construction, and 47 48 equipment. As the main intention of this study was to obtain the operational cost of PCI procedure, 49 land and building costs were not taken into account. Only equipment such as machine, furniture and 50 51 medical instrument with cost higher than RM 1000 and economic useful life exceeded 1 year were 52 53 included. 54 55 56 57 58 59 9 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 10 of 22

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 The complete list of asset in CW and CL, together with their purchase price and year of purchase 4 were obtained from the inventory list. The equivalent annual cost (EAC) was calculated by 5 6 annuitizing the capital outlay using a discount rate with their respective useful life years. Using the 7 8 straight-line depreciation approach, it was assumed that the services from the capital items were 9 divided equally over the useful life of the asset. A discount rate of 3% was chosen in conformity with 10 11 most economic evaluation studies conducted. Based on expert opinion and literature review, a useful 12 13 life of 5 years was used for equipment.[23] The discounted cost of all capital assets was summed up 14 to produce the capital costs. 15 16 For peer review only 17 Consumables 18 19 This included items or goods that are disposable in nature and require regular replacement, such as 20 21 syringes, cotton swabs and needles used during hospitalization. The total costs of general 22 consumables were obtained from the central procurement section, under the purview of pharmacy 23 24 department. 25 26 27 For the purpose of PCI costing, specific consumables referred to those items used in PCI procedure 28 29 such as cardiac stents, catheters, wires and balloons. All the centres have an up-to-date database of 30 the quantity and cost of purchase of these items. 31 32 33 http://bmjopen.bmj.com/ 34 Utility 35 Utility cost referred to expenses incurred on electricity, water, telephone and Internet service. The 36 37 total bill was obtained from the annual financial account. Hospital floor plan from the engineering 38 39 department was obtained to calculate the space ratio, i.e. the percentage of square metres of physical 40 space occupied by the unit of analysis (CW or CL) in comparison to the total hospital indoors area. 41 on October 1, 2021 by guest. Protected copyright. 42 Space ratio was the allocation basis to derive the utility cost of CW and CL respectively. 43 44 45 Hospital Support Service 46 47 For all the participating centres, cost items under the hospital support service such as housekeeping, 48 laundry and linen, waste management, building, and equipment maintenance were outsourced as an 49 50 annual contract to a private company. The annual contract cost for all these services were obtained 51 from the finance department. Similar to utility cost, space ratio was used as the allocation basis to 52 53 derive the hospital support service cost for CW and CL. 54 55 56 57 58 59 10 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 11 of 22 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 Dietary 4 This is the cost of meal preparation by the in-house dietary unit. The total dietary cost was obtained 5 6 from the dietary unit and allocated to CW based on inpatient bed days. 7 8 9 Ancillary Service 10 11 Ancillary services from pharmacy, radiology, laboratory, and rehabilitation departments are usually 12 13 provided to all hospital users. Due to the constraints in time and budget, separate costing analysis 14 was not conducted to assign the cost of each ancillary department to the individual patients. Instead, 15 16 we followed the recommendationFor peer by Hendriks review et al. and grouped only these four ancillary departments as 17 a context-specific overhead category.[24] Its total cost would be the summation of the cost of general 18 19 consumables supplied, the full salary pay-out for all the staff from the four departments, and the 20 21 utility and hospital support cost calculated in the same manner as mentioned above. 22 23 24 Based on a report in Malaysia, 60% of clinical support services such as those provided by the 25 26 ancillary departments can be attributed to the inpatient use, whereas the remaining 40% would be 27 attributed to outpatient use.[25] Therefore, the same 60:40 proportion was applied in this study for 28 29 the cost of ancillary services. Using the ratio of CW inpatient to total hospital inpatient bed days as 30 allocation basis, the cost of ancillary service for CW was calculated. 31 32 33 http://bmjopen.bmj.com/ 34 Step 4 Calculation of Unit Cost 35 36 Upon valuation of each cost item, the summation of costs gave rise to the total direct medical and 37 38 overhead costs at each unit of analysis. This in turn led to the total cost incurred by all patients in 39 40 CW and CL. However, to delineate the cost incurred by patients who underwent elective PCI, 41 suitable denominators must be applied. For CW, the inpatient bed day was deemed as an objective on October 1, 2021 by guest. Protected copyright. 42 43 indicator for the workload. By dividing the total cost of all cost items in with the number of CW 44 inpatient bed days during the study period, we were able to produce the cost of CW per bed day. The 45 46 cost per bed day was multiplied by an average length of stay to produce the cost of CW admission 47 48 per elective PCI. The average length of stay (ALOS) for an elective PCI was decided to be 3 days via 49 consensus by all the interventional cardiologists in the study. 50 51 52 53 For CL, the second unit of analysis, the number of all CL procedures conducted was used as the 54 denominator to derive the cost per CL utilization. As PCI consumables were not shared by all 55 56 procedures conducted in CL, we applied a separate denominator i.e. the total number of PCI 57 58 59 11 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 12 of 22

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 procedures, to calculate the PCI consumables cost per procedure. Together, this gave rise to cost per 4 PCI procedure in CL. 5 6 7 8 The summation of cost per CW admission and cost of PCI procedures in the CL led to the total 9 hospitalization cost of a patient who underwent PCI. It is important to note that the cost produced is 10 11 via a top-down approach and it represented the average cost estimates for a patient admitted for an 12 13 elective PCI procedure, assuming average length of stay of 3 days, with no complications post PCI 14 that required intensive care. This cost also did not reflect any difference in the type and number of 15 16 stents used for each Forpatient. peer review only 17

18 19 RESULTS 20 21 22 Based on the methods described, a range of results can be generated from this study. In this paper, 23 24 we presented the preliminary results from a single centre (Table 2). The ratio of cost incurred at CW 25 26 to CL was 1:4. By using ALOS of 3 days, the cost per admission in Centre I was RM 2,720. This 27 accounted for 19.0% of the total hospitalization cost. Of the CL cost, CL utilization cost per 28 29 procedure was RM 3,821. PCI consumables contributed more than half of the total hospitalization 30 cost. From the breakdown, direct medical cost items were more expensive if compared to overhead 31 32 cost items in both CW and CL. For CW admission cost, medication and labour costs were the 33 http://bmjopen.bmj.com/ 34 highest. On the contrary, capital cost ranked the highest under CL utilization, attributed by the high 35 cost of the cardiac angiography machine. 36 37 38 39 By conducting the same process in the remaining centres, the cost of elective PCI between the 40 centres can be produced. These average cost estimates provide a reflection of the resource utilization 41 on October 1, 2021 by guest. Protected copyright. 42 and cost consumption at the five centres. Comparisons can be made at the cost item level, CW or CL 43 level, up to the overall hospitalization cost. Hospital managers and policy makers will find this 44 45 information beneficial for resource distribution and budget allocation. Furthermore, The unit cost 46 47 estimates generated at cost item level would facilitate any future economic evaluation such as cost 48 effectiveness analysis. It could also be applied for costing of other services in the study centres. 49 50 51 52 53 54 55 56 57 58 59 12 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 13 of 22 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 Table 2 Preliminary Cost Results from Centre I 4 5 Cost Items Of Each Unit Of Analysis Unit Cost 6 Cardiac Ward Admission (CW) 7 Direct Medical 8 Labour 212.87 9 10 Capital 122.18 11 Consumable 37.08 12 Medication 337.47 13 Subtotal 709.60 14 Overhead 15 16 Utility For peer review only27.85 17 Dietary 8.33 18 Hospital Support 26.54 19 Ancillary support 134.36 20 21 Subtotal 197.08 22 Average Cost per Cardiac Ward Bed Day 906.68 23 Average Cost per Admission (ALOS 3 days), A 2,720.04 (19.0%) 24 25 Cardiac Catheterization Laboratory Utilization (CL) 26 27 Direct Medical 28 Labour 810.63 29 Capital 2,619.31 30 31 Consumable 185.8 32 Subtotal 3,615.74 33 Overhead http://bmjopen.bmj.com/ 34 Utility 105.6 35 36 Hospital support 100.64 37 Subtotal 206.25 38 Average Cost per CL utilization, B 3,821.99 (26.7%) 39 40

PCI Consumables on October 1, 2021 by guest. Protected copyright. 41 42 Stents 6,267.07 43 Others 1,499.91 44 Average Cost of PCI Consumables per PCI, C 7,766.98 (54.3%) 45 Average Cost of PCI procedure in CL, B+C 11,588.97 (81.0%) 46 47 48 TOTAL HOSPITALIZATION COST, A+B+C 14,309.01 (100%) 49 50 51 52

53 54 55 56 57 58 59 13 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 14 of 22

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 DISCUSSION 4

5 6 As the leading cause of mortality and morbidity in LMIC, service expansion of the cardiology 7 8 service is inevitable. However, many such countries lack reliable evidence to guide the budget and 9 resource planning. Majority of the LMIC do not have any official health economic evaluation 10 11 guidelines in place, and the international guidelines are not a good fit for LMIC. There is also a lack 12 13 of infrastructure support to sustain a comprehensive health informatics data. These are indispensable 14 components to produce reliable cost-related outcome evidence towards better policy-making. In 15 16 Malaysia specifically,For there is peera lack of easily review accessible information only such as accurate unit costs, up- 17 to-date financial account and complete patient records, all of which are vital to conduct health 18 19 economic evaluation. In this paper, we outlined the study design of a modified costing analysis for 20 21 elective PCI in 5 cardiac centres in Malaysia, based on the limitations in our local setting. We also 22 included the preliminary cost results from one of the centres. 23 24 25 26 We made some modifications to previously published health economics guidelines to adapt them to 27 local setting. Firstly, like many LMIC, casemix system such as DRG is not a common practice in our 28 29 local hospitals. To overcome this, we used an event pathway from admission to discharge to identify 30 the cost items under the respective units of analysis. The lack of DRG and reference cost indirectly 31 32 mandated us to narrow our inclusion criteria to limit the units of analysis. As such, we included only 33 http://bmjopen.bmj.com/ 34 patients who had elective PCI, as they were unlikely to develop post-procedural complications that 35 require admission to intensive care unit. 36 37 38 39 Furthermore, reference cost is also not readily available unlike most of the high-income countries. 40 To reflect the actual cost, certain literature proposed the application of cost-to-charge ratio.[26] This 41 on October 1, 2021 by guest. Protected copyright. 42 method would be relatively simpler and less time consuming. However, it would be a suboptimal 43 cost estimates as public healthcare is highly subsidized in Malaysia and the charges would be a poor 44 45 proxy for the actual cost incurred in the provision of healthcare services. As a result, we conducted 46 47 actual calculation of unit cost for all the cost items with top-down approach. Between top-down and 48 bottom-up approaches, there is no consensus as to which is the gold standard. Typically, top-down 49 50 approach is more straightforward to undertake but the trade-off is a lack of sensitivity especially for 51 cost items that originated from a less homogenous production. On the contrary, bottom-up approach 52 53 is able to produce a more accurate cost estimate, but the main limitation lies within the complexity 54 55 and cost of its implementation, especially across different study centres.[27] Several authors 56 advocated bottom-up approach but acknowledged that such data may not be easily available 57 58 59 14 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 15 of 22 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 especially in resource-limited setting in LMIC.[3, 28, 29] As this costing study involved five centres, 4 a full top-down approach was applied to standardize the cost data collection in order to produce 5 6 reliable cost estimates for comparison between the centres. 7 8 9 There were several limitations. Selection bias existed, as only public cardiac centres were included. 10 11 Due to the sensitive nature of the financial data required, private for-profit healthcare centres were 12 13 reluctant to participate. Despite this, the cost output from public cardiac centres would be of 14 tremendous value for policy makers as the major bulk of CVD cost and workload burden in Malaysia 15 16 fell on the public healthcareFor system. peer This is reflectedreview in a national only survey that showed 65.0% of the 17 Malaysian population sought treatment at public healthcare facilities and 52.4% of the total health 18 19 expenditure was incurred in the public sector.[8, 30] Moreover, potential bias due to inherent 20 21 differences in the financial record keeping between the centres may not be completely eliminated 22 despite our best effort to standardize the data collection process. Despite the shortcomings, this 23 24 costing study across five cardiac centres with different ownership and administration characteristics 25 26 will generate basic cost data that can lead to comprehensive costing analysis. In addition to 27 producing unit cost estimates per output along the clinical pathway, multiple cost components 28 29 involved in the PCI service can be aggregated to identify the dominant cost drivers, both within and 30 across the centres. 31 32 33 http://bmjopen.bmj.com/ 34 Preliminary results from one centre showed the cost incurred in the cardiac catheterization laboratory 35 was four times that of the cost of cardiac ward admission. However, closer scrutiny revealed that 36 37 more than half of the total hospitalization cost could be attributed to PCI consumable. Based on this, 38 39 cost-saving strategy should be focused on the purchasing mechanism of PCI consumable, especially 40 the cardiac stents. To the best of our knowledge, this would be the first multi-centre costing analysis 41 on October 1, 2021 by guest. Protected copyright. 42 of cardiac service in Malaysia. More comprehensive findings can be elicited after consolidating the 43 cost outputs from all centres. Furthermore, the results generated from this costing analysis can be 44 45 incorporated with appropriate clinical data to facilitate individual-level patient analysis, such as the 46 47 identification of cost predictors via regression analysis. The unit cost at multiple level of the care 48 pathway can also be applied to economic evaluation such as cost effectiveness analysis and 49 50 economic modeling purposes in relevant cardiology research. All these will provide valuable 51 information to the cardiac centres and policy makers for proper resource and budget allocation. 52 53 54 55 56 57 58 59 15 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 16 of 22

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 CONCLUSION 4

5 6 The CVD epidemic in LMIC calls for targeted interventions from all aspects of healthcare delivery. 7 8 Given the scarcity of healthcare resources, economic evaluations are needed to identify and 9 implement the most cost effective intervention. Currently, the application of outcomes from 10 11 economic evaluation to health policy decision-making in Malaysia is very limited due to the many 12 13 obstacles in conducting costing studies. As such, we shared the alternative pathways we devised to 14 overcome barriers in conducting a costing study for elective PCI in a LMIC setting. Future studies 15 16 can be built on our effortsFor and peerlike-minded researchersreview in other LMIConly who wish to conduct similar 17 studies can adapt this method to their respective country-specific settings. 18 19 20 21 22 FOOTNOTES 23 24 Acknowledgements 25 26 27 The authors thank all cardiologists and nurses at each participating centres for their effort in data 28 collection. 29 30 31 Contributors 32

33 http://bmjopen.bmj.com/ OI, MD, LHB, OTK, RMA, WAWA, LSY, LEV and KYL conceived the overall study. All authors 34 35 are involved in the design of the study. WAWA, OI, LHB, OTK and RMA led the implementation of 36 37 the study at each centre; while KYL, LEV, LSY, LA and SH coordinated the data collection. KYL 38 drafted the manuscript. TKO and MD revised manuscript critically for intellectual content. All 39 40 authors read and approved the final manuscript.

41 on October 1, 2021 by guest. Protected copyright. 42 43 Funding 44 45 This work was supported by the Malaysia Ministry of Higher Education High Impact Research MoE 46 47 Grant UM.C/625/1/HIR/MoE/ E000010-20001 and Malaysia Ministry of Health MRG grant (MRG- 48 MOH-2014-02). 49 50 51 Competing interests 52 53 All authors declared no competing interest. 54 55 56 57 58 59 16 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 17 of 22 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 Ethics approval 4 5 This study is approved by the Malaysian Ministry of Health Medical Research and Ethics Committee 6 7 (ID: NMRR-13-1403-18234 IIR). 8 9 10 Provenance and peer review 11 12 Not commissioned; externally peer reviewed 13 14 15 Data sharing statement 16 For peer review only 17 No additional data are available 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32

33 http://bmjopen.bmj.com/ 34 35 36 37 38 39 40

41 on October 1, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 17 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 18 of 22

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 REFERENCES 4 1 Drummond M, Sculpher M, Torrance G, et al. Methods For The Economic Evaluation Of Health 5 6 Care Programmes. 3rd ed. Oxford: Oxford University Press, 2005. 7 2 Lipscomb J, Yabroff KR, Brown ML, et al. Health care costing: data, methods, current 8 applications. Medical care 2009;47:S1-6 doi: 10.1097/MLR.0b013e3181a7e401 [published 9 10 Online First: 2009/06/19] 11 3 Mogyorosy Z, Smith PC. The main methodological issues in costing health care services: A 12 literature review. Centre for Health Economics, University of York, 2005. 13 http://www.york.ac.uk/media/che/documents/papers/researchpapers/rp7_Methodological_i 14 15 ssues_in_costing_health_care_services.pdf. 16 4 von Both C, Jahn ForA, Fleba S.peer Costing maternal review health services only in South Tanzania: a case study 17 from Mtwara Urban District. The European journal of health economics : HEPAC : health 18 economics in prevention and care 2008;9:103-15 doi: 10.1007/s10198-007-0048-3 [published 19 20 Online First: 2007/04/25] 21 5 Sarowar MG, Medin E, Gazi R, et al. Calculation of costs of pregnancy- and puerperium-related 22 care: experience from a hospital in a low-income country. Journal of health, population, and 23 24 nutrition 2010;28:264-72 [published Online First: 2010/07/20] 25 6 Briggs A, Nugent R. Editorial. Health economics 2016;25 Suppl 1:6-8 doi: 10.1002/hec.3321 26 [published Online First: 2016/01/26] 27 7 World Health Organization. Global Health Expenditure Database. Global Health Expenditure 28 29 Database 2014. http://apps.who.int/nha/database (accessed 26th July 2016). 30 8 Malaysia National Health Accounts Unit. Malaysia National Health Accounts Health 31 Expediture Report 1997-2012. Malaysia, 2014. 32 9 Planning and Development Division HIC. Health Facts 2013. Malaysia, 2014. 33 http://bmjopen.bmj.com/ 34 http://www.moh.gov.my/images/gallery/publications/HEALTH FACTS 2014.pdf. 35 10 Mendis S. Global status report on noncommunicable disease 2014. 2014. 36 http://www.who.int/nmh/publications/ncd-status-report-2014/en/. 37 11 Nichols M, Townsend N, Scarborough P, et al. Cardiovascular disease in Europe 2014: 38 39 epidemiological update. European heart journal 2014;35:2950-9 doi: 40 10.1093/eurheartj/ehu299 [published Online First: 2014/08/21] 41 12 Townsend N, Bhatnagar P, Wickramasinghe K, et al. Cardiovascular disease statistics, 2014. on October 1, 2021 by guest. Protected copyright. 42 43 London, 2014. 44 13 Lopez A. Global Burden of Disease and Risk Factors. Washington, DC: World Bank 45 Publications, 2006. 46 14 Keeley EC, Boura JA, Grines CL. Primary angioplasty versus intravenous thrombolytic 47 48 therapy for acute myocardial infarction: a quantitative review of 23 randomised trials. Lancet 49 2003;361:13-20 doi: 10.1016/s0140-6736(03)12113-7 [published Online First: 2003/01/09] 50 15 Van de Werf F. Drug-eluting stents in acute myocardial infarction. The New England journal 51 of medicine 2006;355:1169-70 doi: 10.1056/NEJMe068165 [published Online First: 52 53 2006/09/15] 54 16 Negassa A, Monrad ES. Prediction of Length of Stay Following Elective Percutaneous 55 Coronary Intervention. ISRN Surgery 2011;2011:714935 doi: 10.5402/2011/714935 56 57 58 59 18 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 19 of 22 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 17 Shaw RE, Anderson HV, Brindis RG, et al. Development of a risk adjustment mortality model 4 using the American College of Cardiology-National Cardiovascular Data Registry (ACC-NCDR) 5 experience: 1998-2000. Journal of the American College of Cardiology 2002;39:1104-12 6 7 [published Online First: 2002/03/30] 8 18 Batura N, Pulkki-Brannstrom AM, Agrawal P, et al. Collecting and analysing cost data for 9 complex public health trials: reflections on practice. Global health action 2014;7:23257 doi: 10 10.3402/gha.v7.23257 [published Online First: 2014/02/26] 11 12 19 Drummond M, Brandt A, Luce B, et al. Standardizing methodologies for economic evaluation 13 in health care. Practice, problems, and potential. International journal of technology assessment 14 in health care 1993;9:26-36 [published Online First: 1993/01/01] 15 20 Tan SS, Serde ́n L, van Ineveld BM, et al. DRGs and cost accounting: which is driving which? 16 For peer review only 17 Diagnosis-related groups in Europe: towards efficiency and quality. Berlin: Open University 18 Press, 2011. 19 21 NHS reference costs 2014 to 2015. United Kingdom, 2015. 20 21 https://www.gov.uk/government/publications/nhs-reference-costs-2014-to-2015. 22 22 Pharmaceutical Benefits Advisory Committee. Manual of resource items and their 23 associated costs. Canberra, Australia, 2009. 24 http://www.health.gov.au/internet/main/publishing.nsf/Content/health-pbs-general-pubs- 25 26 Manual-content.htm-copy2 27 23 Creese A, Parker D. Cost analysis in primary health care: a training manual for programme 28 managers. Geneva: World Health Organization, 2000. 29 24 Hendriks ME, Kundu P, Boers AC, et al. Step-by-step guideline for disease-specific costing 30 31 studies in low- and middle-income countries: a mixed methodology. Global health action 32 2014;7:23573 doi: 10.3402/gha.v7.23573 [published Online First: 2014/04/02] 33 25 Medical Development Division. Guideline on Cost Data Collection for the Casemix System, http://bmjopen.bmj.com/ 34 35 Malaysian DRG. 2013. 36 26 Shwartz M, Young DW, Siegrist R. The ratio of costs to charges: how good a basis for 37 estimating costs? Inquiry : a journal of medical care organization, provision and financing 38 1995;32:476-81 [published Online First: 1995/01/01] 39 40 27 Emmett D, Forget R. The utilization of activity-based cost accounting in hospitals. Journal of

41 hospital marketing & public relations 2005;15:79-89 doi: 10.1300/J375v15n02_06 [published on October 1, 2021 by guest. Protected copyright. 42 Online First: 2005/10/06] 43 28 Tan SS, Rutten FF, van Ineveld BM, et al. Comparing methodologies for the cost estimation 44 45 of hospital services. The European journal of health economics : HEPAC : health economics in 46 prevention and care 2009;10:39-45 doi: 10.1007/s10198-008-0101-x [published Online First: 47 2008/03/15] 48 49 29 Clement Nee Shrive FM, Ghali WA, Donaldson C, et al. The impact of using different costing 50 methods on the results of an economic evaluation of cardiac care: microcosting vs gross- 51 costing approaches. Health economics 2009;18:377-88 doi: 10.1002/hec.1363 [published 52 Online First: 2008/07/11] 53 54 30 Institute for Public Health. The Third National Health and Morbidity Survey (NHMS III) 55 2006. Kuala Lumpur, 2008. 56 57 58 59 19 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 20 of 22

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 Figure 1 Pathway for the Costing Analysis of Elective PCI 30 31 254x175mm (72 x 72 DPI) 32 33 http://bmjopen.bmj.com/ 34 35 36 37 38 39 40

41 on October 1, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 7-10 7-10 NA NA 1 1

NA NA NA NA 6-7 6-7 5 5

5 5 3-4 3-4

NA NA NA Reported on page # Reported on page 2 2 6 6

BMJ Open http://bmjopen.bmj.com/ on October 1, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

For peer review only ) Indicate study’s) the designcommonlywith used a in term ortitlethe abstract the ) ) theGive eligibility criteria, and sourcesthe methodsand of of selection participants ) ) statisticalDescribe all methods, thoseincluding confoundingusedforcontrol to ) ) Describe any methods used subgroups examine to interactionsand ) If ) applicable, describe methodsanalytical taking account of sampling strategy ) ) in theProvide abstract informativean and balanced summary of what was donewhat wasand found ) Describe) sensitivity any analyses ) how Explain missing data were addressed a a a b e d c b For eachof interest, For variable sourcesdatagive of and details of methods of assessment (measurement). Describe comparability of of comparability assessment methods isthanthere more if one group ( ( why why

applicable collection collection ( ( ( ( Recommendation cross-sectional studies of cross-sectional reports in be included should that items of Statement—Checklist (v4) 2007 STROBE

6 1 ( 3 specific objectives, State including prespecified any hypotheses 5 the setting, Describe locations, relevantand dates, including periods exposure,of recruitment, follow-up, data and 9 any Describe efforts to address sourcespotential of bias 7 define Clearly outcomes,all exposures,predictors, potential confounders,effect and modifiers. diagnostic Give criteria, if 4 key Present of studyelements in thedesign early paper 8* 10 10 how Explain studythe size was arrived at # Item Item Results

Statistical Statistical methods 12 ( Quantitative variables 11 how Explain quantitative were handledvariables in analyses.the If applicable, describe which chosengroupings were and Study size Bias Data sources/Data measurement Variables Participants Setting Study design Methods Objectives Background/rationale 2 thebackground Explain scientific for rationale investigation the and being reported Introduction Title andTitle abstract Section/Topic

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 21 of 22 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from Page 22 of 14-15 14-15 NA NA

15 15 14-15 14-15 14 14

NA NA NA NA NA NA 16 16 NA NA

BMJ Open http://bmjopen.bmj.com/ on October 1, 2021 by guest. Protected copyright. For peer review only For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml ) Report) boundaries category when continuous variables were categorized ) ) unadjustedGive estimates if applicable, and, confounder-adjusted estimates their and precision95% (eg, confidence ) If ) relevant, If consider translating estimatesrelativerisk of into absolute for risk meaningfulperioda time c b a similar studies,similar and other relevant evidence confounders magnitude of magnitude potential biasany confirmed eligible,confirmed included in study, the completing follow-up, analysed and interval). Make interval). clear confounderswhich were adjusted they why and for were included which the presentthe which is based article

( ( number Indicate (b) of participants missingwith eachfor data of interest variable Consider (c) diagram use flow of a reasons Give (b) non-participationfor each at stage 18 18 results Summarise key to reference withstudy objectives 16 16 ( 22 22 sourcethe Give of funding and role the of fundersthe thefor studypresent if and, applicable, originalthe for study on 19 19 Discuss of study, the takinglimitations into sourcesaccount biasof potential imprecision.or Discussboth direction and 13* 13* Report (a) numbers of individuals each at stage of study—eg numbers eligible, potentially examinedfor eligibility, An Explanation and Elaboration article discussesitem each checklist and gives backgroundmethodological and published examples of transparent reporting. STROBE The checklist is best used in this(freelywith conjunction available onarticle sites the Medicine of PLoS Web http://www.plosmedicine.org/,at Annals of Internal Medicine at http://www.annals.org/, http://www.epidem.com/).Epidemiology at and Information on InitiativeSTROBE the is www.strobe-statement.org.available at Note: information*Give separately cases for and controls studiesin forcase-control applicable, if and, exposedunexposed and groups in cohortcross-sectional and studies. Funding Other information Generalisability 21 Discuss(external generalisabilitythe of studythe validity) results Interpretation 20 cautious Give interpretationa of resultsoverall considering limitations, objectives, multiplicity of resultsanalyses, from Limitations Key Key results Discussion Other analyses 17 other Report analyses done—eg analyses of subgroups sensitivity interactions, and and analyses

Main resultsMain Outcome Outcome data 15* numbers Report of outcome events or summary measures

Descriptive data 14* characteristicsGive (a) participantsof study (eg demographic, social) clinical, and information on exposures potential and

Participants 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from

Methods and Rationale of a Multi-Centre Cross Sectional Costing Study of Elective Percutaneous Coronary Intervention: A Study Design based on Malaysia, a Middle- income Country

ForJournal: peerBMJ Open review only

Manuscript ID bmjopen-2016-014307.R1

Article Type: Research

Date Submitted by the Author: 13-Mar-2017

Complete List of Authors: Lee, Kun Yun; University of Malaya, Department of Social and Preventive Medicine Ong, Tiong Kiam; Sarawak Heart Centre, Department of Cardiology Low, Ee Vien; Bahagian Perkhidmatan Farmasi Kementerian Kesihatan Malaysia Liau, Siow Yen; Queen Elizabeth 2 Hospital, Department of Pharmacy, Clinical Research Centre Anchah, Lawrence; Sarawak Heart Centre, Department of Pharmacy Hamzah, Syuhada; Penang General Hospital , Administrative Office Liew, Houng Bang; Queen Elizabeth 2 Hospital, Division of Cardiology, Clinical Research Centre Mohd Ali, Rosli; National Heart Institute, Department of Cardiology http://bmjopen.bmj.com/ Ismail, Omar; Penang General Hospital , Division of Cardiology Wan Ahmad, Wan Azman; Universitiy of Malaya, Division of Cardiology, Department of Medicine, Faculty of Medicine Said, Mas Ayu; University of Malaya, Julius Centre University of Malaya, Department of Social and Preventive Medicine Dahlui, Maznah; University of Malaya, Department of Social and Preventive Medicine

Primary Subject Health economics Heading: on October 1, 2021 by guest. Protected copyright.

Secondary Subject Heading: Cardiovascular medicine, Health services research, Public health

Coronary heart disease < CARDIOLOGY, Coronary intervention < Keywords: CARDIOLOGY, STATISTICS & RESEARCH METHODS, Low- and middle- income countries, Costing approach, Hospitalization Cost

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 1 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 TITLE 4 5 Methods and Rationale of a Multi-Centre Cross Sectional Costing Study of Elective Percutaneous 6 7 Coronary Intervention: A Study Design based on Malaysia, a Middle-income Country. 8 9 10 AUTHORS 11 12 Lee Kun Yun a *, Ong Tiong Kiam b, Low Ee Vien c, Liau Siow Yen d, Lawrence Anchah e, Syuhada 13 f g h i j 14 Hamzah , Liew Houng Bang , Rosli Mohd Ali , Omar Ismail , Wan Azman Wan Ahmad , Mas 15 Ayu Said k, Maznah Dahlui a 16 For peer review only 17 18 a Department of Social and Preventive Medicine, Faculty of Medicine, University of Malaya, 19 Malaysia. 20 21 b Department of Cardiology, Sarawak Heart Centre, Malaysia. 22 23 c 24 Pharmaceutical Services Division, Ministry of Health, Malaysia. 25 d 26 Department of Pharmacy, Clinical Research Centre, Queen Elizabeth 2 Hospital, Sabah, Malaysia. 27 28 e Department of Pharmacy, Sarawak Heart Centre, Malaysia. 29 30 f Administrative Office, Penang General Hospital, Malaysia. 31 32 g Division of Cardiology, Clinical Research Centre, Queen Elizabeth 2 Hospital, Sabah, Malaysia. 33 http://bmjopen.bmj.com/ 34 h Department of Cardiology, National Heart Institute, Kuala Lumpur, Malaysia 35 36 i 37 Division of Cardiology, Penang General Hospital, Penang, Malaysia. 38 j 39 Division of Cardiology, Department of Medicine, Faculty of Medicine, University of Malaya, 40 Malaysia. 41 on October 1, 2021 by guest. Protected copyright. 42 k Julius Centre University of Malaya, Department of Social and Preventive Medicine, Faculty of 43 44 Medicine, University of Malaya, 50603 Kuala Lumpur, Malaysia 45 46 *Corresponding Author 47 48 Email address: [email protected] Telephone number: + 6012-4281182 49 50 51 Postal address: Department of Social and Preventive Medicine, Faculty of Medicine, University of 52 Malaya, Jalan Universtiti, 50603, Kuala Lumpur, Malaysia. 53 54 55 WORD COUNT 4474 56 57 58 59 1 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 2 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 ABSTRACT 4 5 Objectives: 6 7 8 Limitations in the quality and access of cost data from low- and middle-income countries constrain 9 10 the implementation of economic evaluations. With the increasing prevalence of coronary artery 11 disease in Malaysia, cost information is vital for cardiac service expansion. We aim to calculate the 12 13 hospitalization cost of percutaneous coronary intervention (PCI), using a data collection method 14 customized to local setting of limited data availability. 15 16 For peer review only 17 Design: 18 19 This is a cross-sectional costing study from the perspective of healthcare providers, using top-down 20 21 approach, from January-June 2014. Cost items under each unit of analysis involved in the provision 22 of PCI service were identified, valuated and calculated to produce unit cost estimates. 23 24 25 Setting: 26 27 28 Five public cardiac centres participated. All the centres provide full-fledged cardiology services. 29 They are also the tertiary referral centres of their respective regions. 30 31 32 Participants:

33 http://bmjopen.bmj.com/ 34 The cost was calculated for elective PCI procedure in each centre. PCI conducted for 35 36 urgent/emergent indication or for patients with shock and hemodynamic instability were excluded. 37 38 39 Primary and secondary outcome measures: 40 41 The outcome measures of interest were the unit costs at the 2 units of analysis, namely cardiac ward on October 1, 2021 by guest. Protected copyright. 42 43 admission and cardiac catheterization utilization, which made up the total hospitalization cost. 44 45 Results: 46 47 48 The average hospitalization cost ranged between RM11,471 (USD 3,186) and RM 14,465 (USD 49 4,018). PCI consumables were the dominant cost item at all centres. The centre with daycare 50 51 establishment recorded the lowest admission cost and total hospitalization cost. 52 53 54 Conclusions: 55 56 Comprehensive results from all centres enable comparison at the levels of cost items, unit of analysis 57 58 59 2 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 3 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 and total costs. This generates important information on cost variations between centres, thus 4 providing valuable guidance for service planning. Daycare establishment is a feasible cost-saving 5 6 option. For many LMIC with limited data availability, costing method tailored based on country 7 8 setting can be used for the purpose of economic evaluations. 9 10 11 KEYWORDS 12 13 14 15 Costing approach, top-down costing, research methods, hospitalization cost, cardiac centres, low- 16 and middle-income Forcountries peer review only 17 18 19 20 Registration: 21 22 23 24 Malaysian MOH Medical Research and Ethics Committee (ID: NMRR-13-1403-18234 IIR) 25 26 27 28 ARTICLE SUMMARY 29 30 31 Strengths and limitations of this study 32

33 http://bmjopen.bmj.com/ 34 35 • A multi-centre costing analysis using standardized collection methods can lead to within- and 36 37 between-centre comparison at multiple levels, from cost items, units of analysis to overall 38 hospitalization cost. 39 40 • The non-participation of private cardiac centres may limit the generalizability of the results. 41 on October 1, 2021 by guest. Protected copyright. 42 • Top-down costing approach applied in this study produced an average estimate cost per 43 patient and enabled an objective comparison of resource consumption and hospitalization 44 45 cost between different centres. 46 47 • However, this average cost estimates are insufficient for in-depth analysis at patient-level or 48 determination of cost predictors via regression analysis. 49 50 • This alternative costing methods we devised to overcome the data limitations in our setting 51 52 can be adapted by like-minded researchers in other low- and middle-income countries. 53 54 55 56 INTRODUCTION 57 58 59 3 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 4 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 Economic evaluation is an important component of healthcare delivery. With the increasing 4 prevalence of diseases and advancement of medical technologies, healthcare costs continue to 5 6 escalate. In high-income countries (HIC), there are well-established guidelines for healthcare 7 8 economic evaluation.[1] Most of these guidelines were established in the backdrop of an easy access 9 to high quality and freely available data. Therefore, costing analysis of healthcare services are 10 11 commonly applied in HIC to assess the impact of investments in disease prevention and treatment, 12 13 and to guide decision-making in budget allocation and service planning.[2] In contrast, the lack of 14 infrastructure and financial support for evidentiary data capture in low and middle-income countries 15 16 (LMIC) results in a Forpaucity of peereasily accessible review and reliable cost only data. This becomes a huge 17 challenge to the healthcare providers in LMIC towards the uptake of research and application of 18 19 economic evaluation in their daily practice.[3-6] 20 21 22 23 Among the various disease burdens in Malaysia, cardiovascular disease (CVD) is a leading cause of 24 morbidity and mortality. While the CVD mortality rate is declining in developed nations, the drastic 25 26 urbanization and lifestyle changes brought on by epidemiological transition has led to higher 27 28 prevalence of non-communicable diseases in LMIC, with CVD topping the list.[7] As a chronic 29 condition requiring frequent follow-ups and hospitalization, CVD places a huge economic burden on 30 31 the healthcare system and patients alike. Hospitalizations and cost burden attributed to CVD have 32 shown an increasing trend in the European countries.[8-10] Malaysia is not spared of the same CVD 33 http://bmjopen.bmj.com/ 34 epidemic. In 2013, CVD accounted to 24.71% of total mortality in Malaysia and was one of the top 35 36 five causes of hospitalization.[11] 37 38 39 CVD encompasses a range of conditions including coronary artery disease (CAD), cerebrovascular 40

41 disease, peripheral artery disease, and heart failure. Among all, CAD accounts for the highest on October 1, 2021 by guest. Protected copyright. 42 prevalence and mortality. CAD can present as mild angina attack to potentially fatal acute coronary 43 44 syndrome, a spectrum ranging from unstable angina to myocardial infarction. Percutaneous coronary 45 intervention (PCI) is a common treatment modality for CAD due to its safety profile in terms of 46 47 lower mortality and complications.[12, 13] In many countries, the availability of coronary 48 49 catheterization facilities and access to PCI is a key performance indicator of the cardiology service. 50 However, the need for sophisticated laboratory facilities, highly skilled clinical staff, and costly 51 52 consumables such as cardiac stents often drive up the cost of PCI. This can be a constraining factor 53 54 towards its service establishment and delivery in resource-limited countries. 55 56 With the increasing incidence of CAD, expansion of PCI services across Malaysia is underway. 57 58 59 4 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 5 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 More than 60 public, private and teaching institutions, most of which are tertiary-level referral 4 hospitals located in urban areas, are performing approximately 12000 PCIs annually in Malaysia. 5 6 This service is not available in district hospitals of lower tiers. In the public healthcare system, 7 8 cardiology services are provided by teaching hospitals and state-level Ministry of Health hospitals. 9 In addition, Malaysia has a highly specialized national heart institute, which was established as a 10 11 corporatized entity, and given autonomy in terms of financial and staff autonomy, even though it is 12 13 fully owned by the government. Cost of medical procedures and hospitalization often depend on the 14 type of hospitals. Studies have reported cost differences between public- or private-owned hospitals, 15 16 teaching or non-teachingFor institutions. peer Evans reviewet al. reported that averageonly cost of PCI in private 17 teaching hospitals in Japan was much higher than known published records from other hospitals.[14] 18 19 Similar finding was observed between teaching and non-teaching hospitals in the state of New York, 20 21 United States for patients with myocardial infarction.[15] Policy makers and healthcare professionals 22 from other countries will also be interested to gain insight into the local cost variation. 23 24 25 26 PCI and other CVD treatments will continue to evolve with the advancement in medical technology. 27 28 The scarcity of health resources, together with competition for financial budget from other health 29 30 specialties, necessitate the identification of the most cost effective option via economic evaluation. In 31 order to do that, reliable cost data from cardiac centres are needed. However, local evidence are 32

33 lacking; as the development of health economics is still in its infancy stage in Malaysia, and efforts http://bmjopen.bmj.com/ 34 35 to conduct economic evaluations are hampered by limited access to financial data and non- 36 computerized patient clinical data. 37 38 39

40 41 OBJECTIVES on October 1, 2021 by guest. Protected copyright. 42 43 44 With these in mind, we describe a simple, standardized data collection method for a multi-centre 45 46 costing study of cardiac service in Malaysia. Elective PCI is the chosen procedure as it is among the 47 48 most commonly conducted cardiac procedure with the least heterogeneity among the patients and 49 providers. We aim to outline the modifications made to existing guidelines. By presenting the cost 50 51 results from five cardiac centres, we hope to ascertain the cost variation between centres, and to 52 identify potential cost-saving options. 53 54 55 56 METHODS 57 58 59 5 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 6 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 4 Study Design 5 6 7 8 This costing analysis was designed as a hospital-based cross-sectional study, from the perspective of 9 healthcare providers. Cost data collection was conducted from January 1st 2014 to June 30th 2014; 10 11 using a top-down costing approach. All cost estimates were presented in the local currency, 12 13 Malaysian Ringgit (RM), whereby USD 1=RM 3.60 at the time of study. 14 15 16 Patient PopulationFor and Clinical peer Data Collection review only 17

18 19 PCI can be conducted as an emergency or elective procedure. Patients who undergo emergency PCI 20 21 often have more severe comorbidities and disease presentation, leading to higher resource 22 consumption and hospitalization cost. They might develop complications post PCI thus requiring 23 24 admission to an intensive care unit. On the contrary, patients who undergo elective PCI have lower 25 26 risk profile and likely fewer complications. As a relatively homogenous group, patients undergoing 27 elective PCI would be a better proxy to analyze hospital-level cost. As such, we restricted our 28 29 analysis to elective PCI, excluding patients with urgent/emergent indication for PCI, or with shock 30 and hemodynamic instability. This definition is compatible with the definition of elective PCI 31 32 adopted by other studies.[16, 17] 33 http://bmjopen.bmj.com/ 34 35 Study Sites 36 37 38 For the cost data collection, all the public and private cardiac centres were invited to collaborate in 39 40 this study. However, only five public cardiac centres agreed to participate. Being the tertiary referral 41 hospitals of their respective regions, all centres are located in the capital city and most urbanized area on October 1, 2021 by guest. Protected copyright. 42 43 of each region. They provide full-fledged cardiology and cardiothoracic services and is staffed by at 44 least one interventional cardiologist. Centre I is a semi-corporatized university teaching hospital 45 46 while Centre II, III, and IV are government public hospitals. Centre V is a specialized heart centre 47 48 established as a corporatized entity. The range of centres was able to represent possible cost 49 variations stemming from the heterogeneity in hospital characteristics, treatment preferences and 50 51 geographical locations. Table 1 outlines the main description of each centre. 52 53 54 55 56 57 58 59 6 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 7 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 Table 1 Background Information of Study Centres 4 5 6 Centre I II III IV V 7 8 Ownership Type Semi- Public Public Public Corporatized 9 corporatized 10 Year of Initiation of PCI service 1987 1994 2011 2010 1992 11 12 13 *Interventional Cardiologist 5 3 1 1 15 14 15 *Fellow Cardiologists 12 8 7 0 12 16 For peer review only 17 *Total Hospital Admission 34414 29336 1744 9340 8485 18 19 20 *During the study period of January-June 2014 21 22 23 24 25 26 27 28 29 30 31 32

33 http://bmjopen.bmj.com/ 34 35 36 37 38 39 40

41 on October 1, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 7 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 8 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 Cost Data Collection 4 5 6 Due to the inherent characteristics of financial and medical record keeping in each healthcare system, 7 there is no pre-existing data collection tool suitable for use by all countries. For this study, our team 8 9 of health economists and interventional cardiologists created a data collection tool using MS Excel, 10 an easily available programme commonly used in local clinical and research facilities. Figure 1 11 12 outlines the steps of the cost data collection. For each centre, a data enumerator and a site 13 14 coordinator were assigned. The data enumerator was in charge of the data collection from the 15 respective departments. Site coordinators were senior personnel with vast knowledge on the daily 16 For peer review only 17 clinical and financial operations of the centre, especially pertaining to cardiology department and 18 19 PCI procedures. 20 21 22 Pilot study was conducted at Centre IV. Improvements were made based on shortcomings identified 23 and feedbacks from data enumerator, site coordinator and data providers. Training workshops were 24 25 organized, covering topics such as the purpose of the study, definitions of cost items, introduction of 26 27 the data collection tool and instructions on how to use the tool. Upon completion of data collection 28 by data enumerator, site coordinator would check the completeness of the data and conduct simple 29 30 sense check to identify unjustifiable outliers. The compiled cost data was then forwarded to the 31 primary investigator and assessed for its consistency and validity. All unit cost outliers (defined as 32 33 less than one-tenth or more than 10 times the average unit cost at all centres) were flagged and http://bmjopen.bmj.com/ 34 35 queried with the site coordinators. All the cost data were subsequently reviewed in the presence of 36 cardiologists and deemed accurate and justifiable before proceeding on with further calculation and 37 38 analysis. 39 40 41 Costing Pathway (Figure 2) on October 1, 2021 by guest. Protected copyright. 42 43 44 Step 1. Identification of unit of analysis 45 46 47 To achieve the most accurate estimate of PCI cost, it is important to identify all the resources 48 consumed by the patients during hospital admission. For this study, we outlined the event pathway 49 50 for the provision of care for PCI patients from admission to discharge. Two main units of analysis 51 were identified, namely cardiac ward (CW) and cardiac catheterization laboratory (CL). Based on 52 53 guidelines in costing manual, health economists in our study team listed out various cost items that 54 55 should ideally be included under each unit of analysis. The final list of cost items depended on 56 general consensus of interventional cardiologists as they were the subject-matter experts. Under each 57 58 59 8 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 9 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 unit of analysis, cost items that can be directly attributed to patient care are categorized under direct 4 medical cost, while resources that are used by more than one department/unit in the hospital are 5 6 categorized as overhead cost items. 7 8

9 10 Step 2. Identification of Cost Items 11 12 For CW, direct medical cost items were labor, capital, consumables, and medication; whereas 13 14 overhead costs included utility, dietary, ancillary service and hospital support service. There were 15 less cost items under CL, the second unit of analysis. Medications were prescribed and served at 16 For peer review only 17 CW, thus not taken as a cost item under CL. Dietary and ancillary support service were also provided 18 19 only once under CW. 20 21 22 Step 3. Valuation of Cost Items (Table 2) 23 24 25 As reference cost is still under development for public hospitals in Malaysia, actual calculation of 26 unit cost for all the cost items was conducted using a top-down approach in our study. The 27 28 subsequent sections discussed the valuation of each cost item. 29 30 31 Labour 32

33 This referred to the full salary inclusive of wages and employers’ social contributions paid for the http://bmjopen.bmj.com/ 34 staff. A list of all the clinical staff involved in the provision of PCI service were used to obtain their 35 36 payroll register from the finance department. For clinical staff stationed full time in the same unit, 37 38 for example nurses and attendants in CW, laboratory technicians and radiographers in CL, their 39 salaries were summed up in total as all of their productivity were contributed to the units they were 40

41 stationed in. As for doctors, their work scope may spread across several units; including conducting on October 1, 2021 by guest. Protected copyright. 42 43 procedures in the CL, treating patients in the specialist clinic, and leading the ward rounds in CW. 44 The interventional cardiologists in our study team were also the head of cardiology department of 45 46 each centre. Therefore, their expert opinion was sought on the portion of time spent on various 47 activities by doctors of different levels at their own respective centre. The apportioned labour cost of 48 49 doctors based on working hours assumed that all doctors were equally productive in delivering 50 51 patient-related care and no idle time was spent. 52 53 54 Capital 55 56 57 58 59 9 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 10 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 This included fixed, one-time expenses incurred on the purchase of land, building, construction, and 4 equipment. As the main intention of this study was to obtain the operational cost of PCI procedure, 5 6 land and building costs were excluded. Only equipment such as machine, furniture and medical 7 8 instrument with cost higher than RM 1000 and economic useful life exceeding 1 year were included. 9 The complete list of asset in CW and CL, together with their purchase price and year of purchase 10 11 were obtained from the inventory list. Using the straight-line depreciation approach, it was assumed 12 13 that the services from the capital items were divided equally over the useful life of the asset. A 14 discount rate of 3% and a useful life of 5 years were used in conformity with most economic 15 16 evaluations.[18] TheFor equivalent peer annual cost review(EAC) of each asset only was calculated by annuitizing the 17 capital outlay using a discount rate with their respective useful life years. 18 19 20 21 Consumables 22 This included items that are disposable in nature and require regular replacement, such as syringes, 23 24 cotton swabs and needles. The costs of general consumables were obtained from the procurement 25 26 section of pharmacy department. For each of the cardiac centre, consumable provision may occur at 27 different levels, either at the level of individual wards or departments. Depending on the best 28 29 available cost information available, workload ratio from the corresponding unit was used as the 30 allocation basis to calculate the consumable cost. As for PCI-specific consumables such as cardiac 31 32 stents, catheters, wires and balloons, the quantity and cost of purchase were retrieved from the 33 http://bmjopen.bmj.com/ 34 purchase ledger of each cardiology department. 35 36 37 Medication 38 39 Under the drug purchasing mechanism in public hospitals, general medication and cardiac-specific 40 medication were purchased by central pharmacy under separate budget. General medications are 41 on October 1, 2021 by guest. Protected copyright. 42 often stored as floor stock in the ward whereas cardiac-specific medications would be distributed to 43 cardiology department. Appropriate workload ratio was used as the allocation basis to determine the 44 45 cost of general and cardiac-specific medication. 46 47 48 Utility 49 50 Utility cost referred to expenses incurred on electricity, water, telephone and Internet service. The 51 total bill was obtained from the annual financial account. Hospital floor plan from the engineering 52 53 department was obtained to calculate the space ratio, i.e. the percentage of square metres of physical 54 55 space occupied by the unit of analysis (CW or CL) in comparison to the hospital indoors area. Space 56 ratio was the allocation basis to derive the utility cost of CW and CL. 57 58 59 10 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 11 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 4 Hospital Support Service 5 6 For all the participating centres, hospital support services such as housekeeping, laundry, waste 7 8 management, building, and equipment maintenance were outsourced as an annual contract to a 9 private company. The annual contract cost were obtained from the finance department. Similar to 10 11 utility cost, space ratio was used as the allocation basis to derive the hospital support service cost for 12 13 CW and CL. 14 15 16 Dietary For peer review only 17 Cost of meal preparation by the in-house dietary unit was derived by allocating the total dietary cost 18 19 to CW based on inpatient bed days. 20 21 22 Ancillary Service 23 24 Services from pharmacy, radiology, laboratory, and rehabilitation departments are usually provided 25 26 to all hospital users. Due to the constraints in time and budget, separate costing analysis was not 27 conducted to assign the cost of each of these department to the individual patients. Instead, we 28 29 followed the recommendation by Hendriks et al. and grouped them as a context-specific overhead 30 category, labelled as ancillary service.[19] Its total cost would be the summation of the cost of 31 32 general consumables supplied, the full salary pay-out for all the staff, the utility and hospital support 33 http://bmjopen.bmj.com/ 34 services cost calculated in the same manner as mentioned above. 35 36 37 Based on a report in Malaysia, 60% of clinical support services such as those provided by the 38 39 ancillary departments in our study can be attributed to the inpatient use, compared to 40% for 40 outpatient use.[20] Therefore, the same 60:40 proportion was applied in this study for the cost of 41 on October 1, 2021 by guest. Protected copyright. 42 ancillary services. Using the workload ratio of CW to hospital inpatient bed days as allocation basis, 43 the cost of ancillary service was calculated. 44 45 46 47 Step 4 Calculation of Unit Cost 48 49 Upon valuation of each cost item, the summation of costs gave rise to the total direct medical and 50 51 overhead costs incurred by all patients at each unit of analysis. To delineate the cost incurred by 52 53 patients who underwent elective PCI, suitable denominators were applied. For CW, the inpatient bed 54 day was deemed as an objective indicator for the workload. By dividing the total cost of CW with 55 56 the number of CW inpatient bed days during the study period, we were able to produce average cost 57 58 59 11 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 12 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 per bed day. The cost per bed day was multiplied by an average length of stay (ALOS) to produce 4 the average cost per admission. The ALOS for an elective PCI patient was decided to be 3 days via 5 6 consensus by the interventional cardiologists in the study. For Centre III, the weighted admission 7 8 cost was calculated based on the ratio of elective PCI patients admitted to inpatient ward and 9 daycare. 10 11 12 13 For CL, the second unit of analysis, the number of CL procedures conducted was used as the 14 denominator to derive the cost per CL utilization. As PCI consumables were not shared by all 15 16 procedures conductedFor in CL, wepeer applied a separate review denominator only i.e. the number of PCI procedures, 17 to calculate the PCI consumables cost per procedure. Together, this gave rise to cost per PCI 18 19 procedure in CL. 20 21 22 The summation of cost per admission and cost of per PCI procedures in CL led to the total 23 24 hospitalization cost of a patient who underwent PCI. 25 26 27 28 29 30 31 32

33 http://bmjopen.bmj.com/ 34 35 36 37 38 39 40

41 on October 1, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 12 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 13 Total Cost of Cost Items Items Cost of Cost Total Allocation Method for Calculation of of Calculation for Method Allocation Total salary of all staffs staffs of all salary Total Apportioned working hours * salary payout of payout salary * hours working Apportioned doctors Allocation of total cost to workload ratio of CW CW of ratio workload to cost total of Allocation Allocation of total cost to CW or CL to or CL to CW cost total of Allocation ratio workload corresponding consumable inpatient Total= cost consumable inpatient total to days bed CW of* ratio cost days bed hospital inpatient bed days. days. bed inpatient hospital cost to workload ratio of CW bed days over over days bed CW of ratio workload to cost Allocation of total inpatient ancillary services services ancillary inpatient total of Allocation Assumption of 60:40 ratio of ancillary services services of ancillary ratio 60:40 of Assumption CW or CL over total indoor area. area. indoor total over or CL CW Allocation of total contract cost to space ratio of to space cost contract total of Allocation Centre I and and II. I Centre Allocation of total cost to workload ratio for ratio workload to cost total of Allocation was available. available. was and cardiac-specific medication supplied to CW CW to supplied medication cardiac-specific and For Centre III, IV and V, actual cost of general of general cost V, actual and IV III, Centre For Total equivalent annual cost (EAC) of all assets assets all costof (EAC) annual equivalent Total procedure per cost Average Allocation of total utility bill to space ratio of to space bill utility total of Allocation area. indoor total over or CL CW bed days over hospital inpatient bed days days bed inpatient hospital over days bed Eg. For Centre I, allocated CW general general CW I, allocated Centre For Eg. by inpatient versus outpatient. outpatient. inpatient versus by No further allocation needed. needed. allocation further No

(Source)

Not Applicable Applicable Not Not Applicable Applicable Not

BMJ Open All Inpatient departments departments All Inpatient CW/CL CW/CL department Cardiology All Inpatient departments departments All Inpatient CW/CL CW/CL Cardiology department department Cardiology Cost Item Calculation Calculation Item Cost http://bmjopen.bmj.com/ Resource Output Data Required for Required Data Output Resource

ii. iii. i. iii. ii. i. Time spent for different work activities activities work different for spent Time Department) Cardiology of (Head (Patient record office, CW census, CL procedure log) procedure CL census, CW office, record (Patient (Patient record office, CW census) CW office, record (Patient Hospital inpatient bed days days bed inpatient Hospital CW bed days bed CW (Engineering) Space area of hospital indoor area, CW and CL CL and CW area, indoor of hospital area Space (Patient record office, CW census, CL procedure logs) logs) procedure CL census, CW office, record (Patient (based on the best cost data available) dataavailable) cost best the on (based Workload Ratio in terms of inpatient bed days at: days bed inpatient of in terms Ratio Workload

Workload ratio in terms of inpatient bed days at: days bed inpatient of terms in ratio Workload dataavailable) cost best the on (based CW and hospital inpatient bed days days bed inpatient hospital and CW census) CW office, record (Patient Total number of PCI procedures procedures of PCI number Total logs) procedure (CL Space area of hospital indoor area, CW and CL CL and CW area, indoor of hospital area Space (Engineering)

on October 1, 2021 by guest. Protected copyright.

For peer review only For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml (Source) (Finance) (Finance) Cost Items Items Cost All Inpatient departments departments Inpatient All Cardiology department department Cardiology CW/CL CW/CL All Inpatient departments departments Inpatient All CW/CL CW/CL Cardiology department department Cardiology

Cost Data Required for Valuation of of for Valuation Required Data Cost

iii. ii. i. iii. ii. i. (Procurement section of pharmacy department) department) pharmacy of section (Procurement (based on the availability of any of the below) below) of the any of availability the on (based (Finance, Human Resource, Pharmacy, Pharmacy, Resource, Human (Finance, and utility costs based on space ratio space on based costs utility and all four departments, allocated hospital support support hospital allocated departments, four all Labour cost for all personnel, consumables cost to to cost consumables forpersonnel, all cost Labour (Finance) (Finance) Annual contract cost cost contract Annual (Procurement section of pharmacy department) department) pharmacy of section (Procurement (based on the availability of any of the below) below) of the any of availability the on (based Total cost of general medication supplied to: supplied medication of general cost Total

(Engineering, Finance) Finance) (Engineering, Asset inventory, year of purchase, purchase price price purchase of purchase, year inventory, Asset (Human Resource, Finance) Resource, (Human (Human Resource, Finance) Resource, (Human Total cost of catheters, stents, balloons, and wires. wires. and balloons, stents, of catheters, cost Total ledger) purchase department (Cardiology Total dietary cost cost dietary Total (Dietary) Monthly utility bill utility Monthly Engineering) Engineering)

PCI-

Others Others than doctors other forstaffs payout Salary

Doctor Doctor grades different of fordoctors payout Salary specific specific General General to: supplied consumable of general cost Total

Medi- cation cation Consumables Labour Utility Capital Capital Dietary Dietary Support Support Hospital Hospital Ancillary Ancillary Cost Items Items Cost Table 2 Data Required and Allocation Method for the Calculation of Cost Items Items Calculation of Cost the for Method and Allocation Required Table 2 Data 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 13 of 28 BMJ Open Page 14 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 RESULTS 4

5 6 Based on the methods described, a range of results can be generated (Table 3). Comparisons 7 8 can be made at the levels of cost item, unit of analysis (CW or CL), up to the total cost. It is 9 important to note that the cost was produced via a top-down approach and it represented the 10 11 average cost estimates for a patient admitted for an elective PCI procedure, assuming average 12 13 length of stay of 3 days, with no complications post PCI that required intensive care. This 14 cost also did not reflect any difference in the type and number of stents used for each patient. 15 16 For peer review only 17 The total hospitalization cost ranged from RM11,471 (USD 3,186) to RM 14,465 (USD 18 19 4,018). The admission cost accounted for one-fifth or less of the total hospitalization cost at 20 21 all centres. On the contrary, cost incurred at CL was four times that of the admission cost. 22 PCI consumables contributed to the biggest proportion among all the cost items, accounting 23 24 for more than half of the total costs. 25 26 27 Total hospitalization cost was the highest at Centre II. Closer scrutiny revealed that average 28 29 cost of PCI consumables of Centre II, and its proportion over the total hospitalization cost 30 were highest across the 5 centres. In contrast, Centre III, the only centre with a daycare 31 32 service for PCI patients, recorded the lowest cost of CW admission. This partly contributed to 33 http://bmjopen.bmj.com/ 34 cheapest hospitalization cost in Centre III, at 12.3% to 26.1% lower compared to the other 35 centres. Direct medical cost items were generally more expensive than overhead cost items 36 37 in both CW and CL. For CW admission, medication and labour costs were the highest. As for 38 39 CL utilization, capital cost ranked the highest at Centre I and V, the two corporatized 40 hospitals. 41 on October 1, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 14 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 15 1558 5417 2526 830.18 830.18

86,215.03 19,857.02 11,603.16 26,269.15 52,784.55 80,074.16 218,189.52 225,919.31 575,258.55 5,060,957.92 7,238,726.57 2,097,039.64 1,244,978.02 23,005,484.91 10,261,691.59 12,743,046.50 12,743,046.50 4,246.9 (29.9%) 4,246.9 8,179.11 (57.6%) 8,179.11 1,785.56 (12.6%) (12.6%) 1,785.56 14,211.82 (100%) 14,211.82 12,426.01 (87.4%) 12,426.01

251 940 2887 479.59 479.59

15,417.61 70,137.06 47,716.53 62,189.01 62,473.17 43,305.00 500,037.88 782,479.78 334,733.31 345,683.73 250,095.62 111,235.99 461,883.47 1,702,805.64 1,384,582.52 2,152,343.50 2,152,343.50 Centre IV IV Centre V Centre 1,811.5 (14.1%) 1,811.5 8,575.07 (66.6%) 8,575.07 2,490.54 (19.3%) 2,490.54 12,877.11 (100%) 12,877.11 10,386.57 (80.7%) 10,386.57

282 1485 3892 818.53 818.53

49,302.66 87,785.37 64,321.29 58,374.60 Daycare Daycare 778,561.21 687,789.25 398,523.57 193,787.82 558,577.96 124,948.95 222,476.63 622,613.16 2,001,962.06 3,185,726.21 1,340,625.80 2,379,464.50 2,379,464.50

9,785.95 (85.3%) 9,785.95 8,437.82 (73.6%) 8,437.82 1,348.13 (11.8%) (11.8%) 1,348.13 11,471.04 (100%) 11,471.04 *1,685.09 (14.7%) *1,685.09 Centre III Centre 2786 647.87 647.87

94,791.64 97,277.64 95,143.27 35,375.00 874,136.64 214,708.34 287,242.14 106,302.47 Inpatient Inpatient 1,804,977.14

257 1118 3682 595.19 595.19

22,559.11 34,634.59 63,584.90 12,697.27 880,833.45 614,370.96 359,953.93 222,997.80 154,201.25 118,721.94 159,664.13 126,005.76 BMJ Open 1,912,352.04 2,191,477.81 1,333,604.78

2,778,554.00 2,778,554.00 Centre II Centre http://bmjopen.bmj.com/ 1,710.51 (11.8%) 1,710.51 1,943.63 (13.4%) 1,943.63 14,465.64 (100%) 14,465.64 12,522.00 (86.6%) 12,522.00 10,811.49 (74.7%) 10,811.49

580 1084 3129 906.68 906.68

83,040.43 87,132.83 26,075.00 420,404.41 201,404.15 116,035.54 109,096.89 878,721.52 114,473.40 666,064.92 382,311.39 1,055,939.18 2,839,335.96 4,143,031.92 2,837,003.70 4,504,850.00 4,504,850.00 7,766.98 (54.3%) 7,766.98 3,821.99 (26.7%) 3,821.99 2,720.04 (19.0%) 2,720.04 on October 1, 2021 by guest. Protected copyright. 14,309.01 (100%) 14,309.01 11,588.97 (81.0%) 11,588.97

For peer review only For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

Total Cost of PCI Consumables Consumables of PCI Cost Total Total Cost of CL of CL Cost Total Labour Labour Capital Consumable Utility Support Hospital Average Cost per Bed Day Bed per Cost Average Total cost of CW ofCW cost Total Labour Labour Capital Consumable Medication Utility Dietary Support Hospital support Ancillary Number of PCI Procedures Procedures PCI of Number Number of CL Procedures Procedures CL of Number Number of CW Bed Days Days Bed CW of Number Medical head Medical

Direct Direct Over- Overhead Overhead Direct Direct

*Weighted admission cost was calculated for Centre III based on the ratio of elective PCI patients admitted to inpatient ward versus daycare. daycare. versus ward inpatient to admitted patients PCI elective of ratio on the IIIbased Centre for calculated was cost admission *Weighted Total Hospitalization Cost, A+B+C A+B+C Cost, Hospitalization Total Cardiac Catheterization Laboratory Utilization (CL) (CL) Utilization Laboratory Catheterization Cardiac Cardiac Ward Admission (CW) (CW) Admission Ward Cardiac Table 3 Cost Items of Each Unit of Analysis Contributing to Total Hospitalization Cost of Elective PCI at All Centres All Centres PCI Cost at of Elective to Hospitalization Total Contributing of Analysis Unit of Each Items Table 3 Cost Ringgit) (Malaysian Item ofCost Cost Total I Centre Average Cost of PCI procedure in CL, B+C B+C in CL, procedure ofPCI Cost Average Average Cost of PCI Consumables, C Consumables, ofPCI Cost Average Average Cost per CL utilization, B B utilization, CL per Cost Average Average Cost per Admission, A A Admission, per Cost Average 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 15 of 28 BMJ Open Page 16 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 DISCUSSION 4

5 6 As CVD represents the leading cause of mortality and morbidity in LMIC, service expansion of the 7 8 cardiology service is inevitable. However, most LMIC lack reliable evidence to guide the budget and 9 resource planning as they do not have any official economic evaluation guidelines in place, and the 10 11 international guidelines are not a good fit for LMIC. There is also a lack of infrastructure support to 12 13 sustain a comprehensive health informatics data. These are indispensable components to produce 14 reliable cost-related outcome evidence towards better policy-making. In Malaysia, specifically, there 15 16 is a lack of easily accessibleFor information peer such review as accurate unit costs, only detailed financial account and 17 complete patient records, all of which are vital to conduct economic evaluation. In this paper, we 18 19 outlined the study design of a multi-centre costing analysis for elective PCI, based on the limitations 20 21 in our local setting, before presenting the cost output. Overall, the results demonstrated substantial 22 costs associated with the provision of PCI, with PCI consumables being the dominant cost item 23 24 across all centres. 25 26 27 In the course of designing the costing analysis, we made some modifications to previously published 28 29 guidelines. International guidelines recommended the design of a standardized cost collection tool 30 that is universally available and easy to use for all involved parties.[21, 22] Thus, we chose MS 31 32 Excel over other more sophisticated costing software, in view of its wide accessibility and user- 33 http://bmjopen.bmj.com/ 34 friendliness. Many HIC used Diagnosis Related Group (DRG), a system which classifies acute 35 inpatient episodes based on the main clinical condition, as unit of analysis for illness-specific 36 37 costing.[23] However, casemix system such as DRG is not a common practice in our local hospitals. 38 39 To overcome this, we used an event pathway from admission to discharge to identify cost items 40 under the respective units of analysis. Furthermore, in HIC, reference cost from national database are 41 on October 1, 2021 by guest. Protected copyright. 42 used for cost item valuation, such as in the United Kingdom and Australia.[24, 25] In the absence of 43 reference cost, certain literature proposed the application of cost-to-charge ratio, which is a relatively 44 45 simpler and less time consuming method to reflect the actual cost.[26] However, it would be a 46 47 suboptimal cost estimate as public healthcare is highly subsidized in Malaysia and the charges would 48 be a poor proxy for the actual cost incurred. As a result, we conducted actual calculation of unit cost 49 50 with top-down approach. Between top-down and bottom-up approaches, there is no consensus as to 51 which is the gold standard. Typically, top-down approach is more straightforward to undertake but 52 53 the trade-off is a lack of sensitivity especially for cost items that originated from a less homogenous 54 55 production. On the contrary, bottom-up approach is able to produce a more accurate cost estimate, 56 but the main limitation lies within the complexity of its implementation, especially across different 57 58 59 16 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 17 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 study centres.[27] Several authors advocated bottom-up approach but acknowledged that such data 4 may not be easily available in resource-limited LMIC setting.[3, 28, 29] As this costing study 5 6 involved five centres, a full top-down approach was applied to standardize the cost data collection in 7 8 order to produce reliable cost estimates for comparison between the centres. 9 10 11 The application of economic evaluation in routine clinical practice is not widespread in Malaysia. 12 13 The lack of familiarity among the staff in healthcare facilities towards the conduct of costing analysis 14 was a major barrier in our study. All private hospitals cited confidentiality issue of financial data 15 16 sharing as reason ofFor non-participation. peer During review the course of data only collection, there were also 17 circumstances of bureaucratic resistance towards parting with cost data of sensitive nature. 18 19 Misperception of the costing exercise as a performance audit by data providers at some hospital 20 21 departments was another reason behind their reluctance to cooperate. As a result, the relationship 22 between the enumerators and other hospital staff proved to be a vital factor towards a successful data 23 24 collection. In several instances, site coordinators had to step in and used their discretion to ensure 25 26 that the staff provided complete data. This highlighted the importance of ensuring access of 27 transparent and quality cost data towards the successful implementation of a costing analysis. 28 29 30 There were several limitations. Selection bias existed, as only public cardiac centres were included 31 32 and there was no input from private for-profit healthcare centres. Despite this, the cost output would 33 http://bmjopen.bmj.com/ 34 be of tremendous value for policy makers as the major bulk of care burden in Malaysia fell on the 35 public healthcare system. This is reflected in a national survey that showed 65.0% of the Malaysian 36 37 population sought treatment at public healthcare facilities and 52.4% of the total health expenditure 38 39 was incurred in the public sector.[30, 31] The exclusion of land and building costs may reduce the 40 applicability of capital cost findings to rural, district hospitals. However, this has less impact towards 41 on October 1, 2021 by guest. Protected copyright. 42 costing of highly specialized medical procedures such as PCI, as the provision of such services in 43 Malaysia will likely remain in tertiary-level hospitals in urban locations. Another limitation being the 44 45 potential bias arising from inherent differences in the financial record keeping between the centres. 46 47 This may not be completely eliminated despite our best effort to standardize the data collection 48 process, as shown in the derivation of consumable and medication costs. Despite the shortcomings, 49 50 the key strengths of this study lie in the practical applications of its methodological contribution and 51 cost findings. We shared the alternative pathways we devised to overcome barriers in conducting a 52 53 costing analysis in a LMIC setting. To the best of our knowledge, this is the first multi-centre costing 54 55 analysis of cardiac service in Malaysia. Average cost estimates derived from five cardiac centres 56 57 58 59 17 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 18 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 with different ownership and administration characteristics enabled us to compare the resource 4 utilization and cost output at various levels of the costing pathway. 5 6 7 8 While direct cost comparison with other published literatures would not be completely feasible 9 owing to the heterogeneous study populations, different perspectives used in cost estimation, varying 10 11 cost calculation methods and vastly different healthcare systems, our finding of PCI consumables 12 13 being the dominant cost items strikes a similar chord with many studies.[32-34] This shows the 14 importance of an efficient consumable purchasing mechanism, especially for cardiac stents. Policy 15 16 makers should reviewFor the current peer procurement review practice of PCI consumablesonly by individual cardiac 17 centres. Vigorous negotiation via competitive bidding processes, ideally handled via a central 18 19 purchasing agency, may lead to potential cost reduction. Furthermore, our results showed that 20 21 daycare establishment can be an attractive cost-saving strategy. The general consensus among 22 studies that looked at the safety of same-day discharge for PCI patients found that low-risk patients 23 24 were good candidates for same-day discharge after uncomplicated PCI.[35, 36] Nevertheless, further 25 26 research in more hospitals need to be carried out to study the feasibility of establishing large-scale 27 daycare service for cardiac patients. 28 29 30 Overall, the corporatization and teaching status of hospitals in this study appear to have minor 31 32 influence to the eventual cost. However, the cost differences may also be due to the casemix of 33 http://bmjopen.bmj.com/ 34 patients treated, and the outcomes of procedures at each centre. It would be interesting if further 35 casemix-adjusted analysis could be conducted to analyze if any of these contributes to the difference 36 37 in the cost output. Furthermore, results from this study can be incorporated with clinical data to 38 39 facilitate individual-level patient analysis, such as the identification of cost predictors via regression 40 analysis. The unit cost derived from the various levels of the care pathway can also be applied to 41 on October 1, 2021 by guest. Protected copyright. 42 economic evaluation such as cost effectiveness analysis or economic modeling in relevant research. 43 All these will provide valuable information to the cardiac centres and policy makers for proper 44 45 resource and budget allocation. Lastly, while in-hospital treatment may account for a substantial 46 47 portion of overall cost of CAD care, pre- and post-PCI cost to the healthcare providers, patients and 48 society may represent hidden economic burden. This is an important area for future research. 49 50 51 52 53 54 55 56 CONCLUSION 57 58 59 18 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 19 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 4 The CVD epidemic in LMIC calls for targeted interventions from all aspects of healthcare delivery. 5 6 Currently, the application of evidence from economic evaluation for health policy decision-making 7 8 in Malaysia is very limited due to the obstacles in conducting costing studies. We devised alternative 9 costing pathways to overcome the barriers and conducted a multi-centre costing analysis for elective 10 11 PCI. The findings highlighted the need for effective procurement practice in view of the high cost of 12 13 PCI consumables. Shorter hospitalization stay via daycare establishment represents another potential 14 cost-saving mechanism. Future studies can be built on our efforts for other medical procedures or 15 16 healthcare service. RecommendationsFor peer from thisreview study would also only be useful for like-minded 17 researchers who wish to conduct similar costing studies in LMIC. 18 19 20 21 22 FOOTNOTES 23 24 Acknowledgements 25 26 27 The authors thank all cardiologists and nurses at each participating centres for their effort in data 28 collection. 29 30 31 Contributors 32

33 http://bmjopen.bmj.com/ OI, MD, LHB, OTK, RMA, WAWA, LSY, LEV and KYL conceived the overall study. All authors 34 35 are involved in the design of the study. WAWA, OI, LHB, OTK and RMA led the implementation of 36 37 the study at each centre; while KYL, LEV, LSY, LA and SH coordinated the data collection. KYL 38 drafted the manuscript. TKO and MD revised manuscript critically for intellectual content. All 39 40 authors read and approved the final manuscript.

41 on October 1, 2021 by guest. Protected copyright. 42 43 Funding 44 45 This work was supported by the Malaysia Ministry of Higher Education High Impact Research MoE 46 47 Grant UM.C/625/1/HIR/MoE/ E000010-20001 and Malaysia Ministry of Health MRG grant (MRG- 48 MOH-2014-02). 49 50 51 52 53 Competing interests 54 55 56 All authors declared no competing interest. 57 58 59 19 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 20 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 Ethics approval 4 5 This study is approved by the Malaysian Ministry of Health Medical Research and Ethics Committee 6 7 (ID: NMRR-13-1403-18234 IIR). 8 9 10 Provenance and peer review 11 12 Not commissioned; externally peer reviewed 13 14 15 Data sharing statement 16 For peer review only 17 No additional data are available 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32

33 http://bmjopen.bmj.com/ 34 35 36 37 38 39 40

41 on October 1, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 20 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 21 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 REFERENCES 4 5 1 Drummond M, Sculpher M, Torrance G, et al. Methods For The Economic Evaluation Of Health 6 Care Programmes. 3rd ed. Oxford: Oxford University Press, 2005. 7 2 Lipscomb J, Yabroff KR, Brown ML, et al. Health care costing: data, methods, current 8 applications. Med Care 2009;47:S1-6 doi: 10.1097/MLR.0b013e3181a7e401 [published Online 9 First: 2009/06/19] 10 3 Mogyorosy Z, Smith PC. The main methodological issues in costing health care services: A 11 literature review. Centre for Health Economics, University of York, 2005. 12 http://www.york.ac.uk/media/che/documents/papers/researchpapers/rp7_Methodological_i 13 ssues_in_costing_health_care_services.pdf. 14 4 von Both C, Jahn A, Fleba S. Costing maternal health services in South Tanzania: a case study 15 from Mtwara Urban District. The European journal of health economics : HEPAC : health 16 For peer review only 17 economics in prevention and care 2008;9:103-15 doi: 10.1007/s10198-007-0048-3 [published 18 Online First: 2007/04/25] 19 5 Sarowar MG, Medin E, Gazi R, et al. Calculation of costs of pregnancy- and puerperium-related 20 care: experience from a hospital in a low-income country. Journal of health, population, and 21 nutrition 2010;28:264-72 [published Online First: 2010/07/20] 22 6 Briggs A, Nugent R. Editorial. Health economics 2016;25 Suppl 1:6-8 doi: 10.1002/hec.3321 23 [published Online First: 2016/01/26] 24 7 Mendis S. Global status report on noncommunicable disease 2014. 2014. 25 http://www.who.int/nmh/publications/ncd-status-report-2014/en/. 26 27 8 Nichols M, Townsend N, Scarborough P, et al. Cardiovascular disease in Europe 2014: 28 epidemiological update. European heart journal 2014;35:2950-9 doi: 29 10.1093/eurheartj/ehu299 [published Online First: 2014/08/21] 30 9 Townsend N, Bhatnagar P, Wickramasinghe K, et al. Cardiovascular disease statistics, 2014. 31 London, 2014. 32 10 Lopez A. Global Burden of Disease and Risk Factors. Washington, DC: World Bank 33 Publications, 2006. http://bmjopen.bmj.com/ 34 11 Planning and Development Division HIC. Health Facts 2013. Malaysia, 2014. 35 http://www.moh.gov.my/images/gallery/publications/HEALTH FACTS 2014.pdf. 36 12 Keeley EC, Boura JA, Grines CL. Primary angioplasty versus intravenous thrombolytic 37 38 therapy for acute myocardial infarction: a quantitative review of 23 randomised trials. Lancet 39 2003;361:13-20 doi: 10.1016/s0140-6736(03)12113-7 [published Online First: 2003/01/09] 40 13 Van de Werf F. Drug-eluting stents in acute myocardial infarction. The New England journal

41 of medicine 2006;355:1169-70 doi: 10.1056/NEJMe068165 [published Online First: on October 1, 2021 by guest. Protected copyright. 42 2006/09/15] 43 14 Evans E, Imanaka Y, Sekimoto M, et al. Risk adjusted resource utilization for AMI patients 44 treated in Japanese hospitals. Health Econ 2007;16:347-59 doi: 10.1002/hec.1177 [published 45 Online First: 2006/10/13] 46 15 Polanczyk CA, Lane A, Coburn M, et al. Hospital outcomes in major teaching, minor teaching, 47 and nonteaching hospitals in New York state. Am J Med 2002;112:255-61 doi: 10.1016/S0002- 48 49 9343(01)01112-3 [published Online First: 2002/03/15] 50 16 Negassa A, Monrad ES. Prediction of Length of Stay Following Elective Percutaneous 51 Coronary Intervention. ISRN Surgery 2011;2011:714935 doi: 10.5402/2011/714935 52 17 Shaw RE, Anderson HV, Brindis RG, et al. Development of a risk adjustment mortality model 53 using the American College of Cardiology-National Cardiovascular Data Registry (ACC-NCDR) 54 experience: 1998-2000. Journal of the American College of Cardiology 2002;39:1104-12 55 [published Online First: 2002/03/30] 56 57 58 59 21 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 22 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 18 Creese A, Parker D. Cost analysis in primary health care: a training manual for programme 4 managers. Geneva: World Health Organization, 2000. 5 19 Hendriks ME, Kundu P, Boers AC, et al. Step-by-step guideline for disease-specific costing 6 studies in low- and middle-income countries: a mixed methodology. Global health action 7 2014;7:23573 doi: 10.3402/gha.v7.23573 [published Online First: 2014/04/02] 8 9 20 Medical Development Division. Guideline on Cost Data Collection for the Casemix System, 10 Malaysian DRG. 2013. 11 21 Batura N, Pulkki-Brannstrom AM, Agrawal P, et al. Collecting and analysing cost data for 12 complex public health trials: reflections on practice. Global health action 2014;7:23257 doi: 13 10.3402/gha.v7.23257 [published Online First: 2014/02/26] 14 22 Drummond M, Brandt A, Luce B, et al. Standardizing methodologies for economic evaluation 15 in health care. Practice, problems, and potential. International journal of technology assessment 16 in health care 1993;9:26-36For peer[published Online review First: 1993/01/01] only 17 23 Tan SS, Serde ́n L, van Ineveld BM, et al. DRGs and cost accounting: which is driving which? 18 Diagnosis-related groups in Europe: towards efficiency and quality. Berlin: Open University 19 20 Press, 2011. 21 24 NHS reference costs 2014 to 2015. United Kingdom, 2015. 22 https://www.gov.uk/government/publications/nhs-reference-costs-2014-to-2015. 23 25 Pharmaceutical Benefits Advisory Committee. Manual of resource items and their 24 associated costs. Canberra, Australia, 2009. 25 http://www.health.gov.au/internet/main/publishing.nsf/Content/health-pbs-general-pubs- 26 Manual-content.htm-copy2 27 26 Shwartz M, Young DW, Siegrist R. The ratio of costs to charges: how good a basis for 28 estimating costs? Inquiry : a journal of medical care organization, provision and financing 29 1995;32:476-81 [published Online First: 1995/01/01] 30 31 27 Emmett D, Forget R. The utilization of activity-based cost accounting in hospitals. Journal of 32 hospital marketing & public relations 2005;15:79-89 doi: 10.1300/J375v15n02_06 [published

33 Online First: 2005/10/06] http://bmjopen.bmj.com/ 34 28 Tan SS, Rutten FF, van Ineveld BM, et al. Comparing methodologies for the cost estimation 35 of hospital services. The European journal of health economics : HEPAC : health economics in 36 prevention and care 2009;10:39-45 doi: 10.1007/s10198-008-0101-x [published Online First: 37 2008/03/15] 38 29 Clement Nee Shrive FM, Ghali WA, Donaldson C, et al. The impact of using different costing 39 methods on the results of an economic evaluation of cardiac care: microcosting vs gross- 40

costing approaches. Health economics 2009;18:377-88 doi: 10.1002/hec.1363 [published on October 1, 2021 by guest. Protected copyright. 41 42 Online First: 2008/07/11] 43 30 Malaysia National Health Accounts Unit. Malaysia National Health Accounts Health 44 Expediture Report 1997-2012. Malaysia, 2014. 45 31 Institute for Public Health. The Third National Health and Morbidity Survey (NHMS III) 46 2006. Kuala Lumpur, 2008. 47 32 Varani E, Balducelli M, Vecchi G, et al. Comparison of multiple drug-eluting stent 48 percutaneous coronary intervention and surgical revascularization in patients with multivessel 49 coronary artery disease: one-year clinical results and total treatment costs. J Invasive Cardiol 50 2007;19:469-75 [published Online First: 2007/11/08] 51 33 Manari A, Costa E, Scivales A, et al. Economic appraisal of the angioplasty procedures 52 53 performed in 2004 in a high-volume diagnostic and interventional cardiology unit. J Cardiovasc 54 Med (Hagerstown) 2007;8:792-8 doi: 10.2459/JCM.0b013e328012c1b6 [published Online 55 First: 2007/09/22] 56 57 58 59 22 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 23 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 34 Chino M, Sakamoto M, Sasaki T, et al. [What aspects are inefficient concerning percutaneous 4 coronary intervention in Japan?: International cost-effectiveness comparisons]. J Cardiol 5 2004;44:85-92 [published Online First: 2004/10/27] 6 35 Abdelaal E, Rao SV, Gilchrist IC, et al. Same-day discharge compared with overnight 7 hospitalization after uncomplicated percutaneous coronary intervention: a systematic review 8 and meta-analysis. JACC Cardiovasc Interv 2013;6:99-112 doi: 10.1016/j.jcin.2012.10.008 9 [published Online First: 2013/01/29] 10 11 36 Brayton KM, Patel VG, Stave C, et al. Same-Day Discharge After Percutaneous Coronary 12 InterventionA Meta-Analysis. Journal of the American College of Cardiology 2013;62:275-85 13 doi: 10.1016/j.jacc.2013.03.051 14 15 16 Figure 1 Pathway forFor the Costing peer Analysis ofreview Elective PCI only 17 18 19 Figure 2 Flow Chart of Task and Person-In-Charge Involved at Each Step of Data Collection Process 20 21 22 23 24 25 26 27 28 29 30 31 32

33 http://bmjopen.bmj.com/ 34 35 36 37 38 39 40

41 on October 1, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 23 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 24 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 Figure 1 Pathway for the Costing Analysis of Elective PCI 30 31 32 137x95mm (300 x 300 DPI) 33 http://bmjopen.bmj.com/ 34 35 36 37 38 39 40

41 on October 1, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 25 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32

33 http://bmjopen.bmj.com/ 34 35 36 37 38 39 40

41 on October 1, 2021 by guest. Protected copyright. 42 43 44 45 Figure 2 Flow Chart of Task and Person-In-Charge Involved at Each Step of Data Collection Process 46 47 48 296x419mm (300 x 300 DPI) 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from Page 26 of 28

Additional Additional remarks

, , Study Sections Study Methods-Costing Methods-Costing Introduction Discussion Discussion Methods Methods Introduction, Methods Methods Introduction, Introduction, Methods Methods Introduction, Introduction Introduction Introduction, Methods Methods Introduction,

Not relevant Not relevant Not applicable applicable Not Not applicable applicable Not Not relevant Not relevant

BMJ Open

http://bmjopen.bmj.com/

on October 1, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

For peer review only

(11) (11) The outcome primary measure(s) the for areevaluation economic (10) (10) Details the of method synthesis of or estimatesof meta- analysis are given based (if an on a overview of ofnumber studies) effectiveness (9) (9) Details the of design and resultsstudy effectiveness of given are (if (8) (8) source(s) The effectiveness of estimates stated used are Data collection collection Data (7) (7) The economic of evaluationchoice form of justified is in to relation the questions addressed (6) (6) The economic of evaluationform used stated is (5) (5) The being alternatives compared are clearly described (4) (4) The choosing rationale for the alternative programmes or interventions compared stated is (3) (3) The viewpoint(s) the of stated are analysis clearly justified and (2) (2) importance The economic the of question research stated is (1) (1) The research statedisquestion Study design design Study based on singlebased study) on 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from

Costing Costing -

Methods-Cost Data Data Methods-Cost 2 Collection-Table

Methods Pathway Pathway Methods-Cost Data Data Methods-Cost and 2 Collection-Table 3 Methods-Study Design Design Methods-Study Methods-Costing Methods-Costing Pathway Not Applicable Not Applicable Not applicable applicable Not Not applicable applicable Not Not applicable applicable Not Not Applicable Not Applicable

BMJ Open

http://bmjopen.bmj.com/

on October 1, 2021 by guest. Protected copyright.

For peer review only For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

Analysis and interpretation of results of results interpretation and Analysis (21) (21) The choice model of used parametersthe and key which iton is (20) (20) Details model any of given used are (19) (19) Details currency of price of adjustments or inflation currency for conversion are given (18) (18) Currency and price data recorded are (17) (17) Methods the quantitiesfor of estimation unit costs are and described (16) (16) Quantities resourcesof separatelyreported are their from unit costs (15) (15) The productivity relevance of changes to the question study is discussed (14) (14) Productivity changes included) separately(if are reported (13) (13) Detailssubjects the whom of valuations from were obtained are given (12) (12) Methods value health statesto benefits other and stated are clearly clearly stated based are justified 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 27 of 28 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from Page 28 of

Only for calculation of calculation Only for Asset under Labour Cost Cost Asset under Labour Only for calculation of calculation Only for Cost Asset under Labour Only for calculation of calculation Only for Cost Asset under Labour

Cost Data Data Cost - Methods Collection-Labour Collection-Labour Methods-Cost Data Data Methods-Cost Collection-Labour Conclusion Conclusion Methods-Cost Data Data Methods-Cost Collection-Labour Methods-Study Design Design Methods-Study Results Results Discussion, Discussion, Conclusion Results, Discussion Results, Discussion Results, Results, Table 3 Not applicable applicable Not Not applicable applicable Not Not applicable applicable Not Not applicable applicable Not Not applicable applicable Not BMJ Open

http://bmjopen.bmj.com/

on October 1, 2021 by guest. Protected copyright.

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

For peer review only

22) Time horizon horizon Time 22) of costs benefits and stated is (35) (35) Conclusions by are the accompanied caveats appropriate (34) (34) Conclusions the follow from data reported (33) (33) The study the answer to given is question (32) (32) Major outcomes are in presented well as disaggregated a as aggregated form (31) (31) Incremental analysis reported is (30) (30) Relevant alternatives are compared (29) (29) The which theranges over variables are are statedvaried (28) (28) The choice variablesof sensitivity for analysis justified is (27) (27) The to approach sensitivity analysis is given (26) (26) Detailsstatistical of tests confidence and intervalsgiven are for stochastic data (25) (25) An given costsexplanation is if are benefits or discounted not (24) (24) The choice rate(s)of is justified ( (23) The rate(s)discount stated is 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from

Cost of Elective Percutaneous Coronary Intervention in Malaysia, a Multi-Centre Cross Sectional Costing Study

Journal: BMJ Open ManuscriptFor ID peerbmjopen-2016-014307.R2 review only Article Type: Research

Date Submitted by the Author: 06-Apr-2017

Complete List of Authors: Lee, Kun Yun; University of Malaya, Department of Social and Preventive Medicine Ong, Tiong Kiam; Sarawak Heart Centre, Department of Cardiology Low, Ee Vien; Bahagian Perkhidmatan Farmasi Kementerian Kesihatan Malaysia Liau, Siow Yen; Queen Elizabeth 2 Hospital, Department of Pharmacy, Clinical Research Centre Anchah, Lawrence; Sarawak Heart Centre, Department of Pharmacy Hamzah, Syuhada; Penang General Hospital , Administrative Office Liew, Houng Bang; Queen Elizabeth 2 Hospital, Division of Cardiology, Clinical Research Centre Mohd Ali, Rosli; National Heart Institute, Department of Cardiology Ismail, Omar; Penang General Hospital , Division of Cardiology Wan Ahmad, Wan Azman; Universitiy of Malaya, Division of Cardiology,

Department of Medicine, Faculty of Medicine http://bmjopen.bmj.com/ Said, Mas Ayu; University of Malaya, Julius Centre University of Malaya, Department of Social and Preventive Medicine Dahlui, Maznah; University of Malaya, Department of Social and Preventive Medicine

Primary Subject Health economics Heading:

Secondary Subject Heading: Cardiovascular medicine, Health services research, Public health on October 1, 2021 by guest. Protected copyright. Coronary heart disease < CARDIOLOGY, Coronary intervention < Keywords: CARDIOLOGY, STATISTICS & RESEARCH METHODS, Low- and middle- income countries, Costing approach, Hospitalization Cost

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 1 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 TITLE 4 5 Cost of Elective Percutaneous Coronary Intervention in Malaysia, a Multi-Centre Cross Sectional 6 7 Costing Study 8 9 10 AUTHORS 11 12 Lee Kun Yun a *, Ong Tiong Kiam b, Low Ee Vien c, Liau Siow Yen d, Lawrence Anchah e, Syuhada 13 f g h i j 14 Hamzah , Liew Houng Bang , Rosli Mohd Ali , Omar Ismail , Wan Azman Wan Ahmad , Mas 15 Ayu Said k, Maznah Dahlui a 16 For peer review only 17 18 a Department of Social and Preventive Medicine, Faculty of Medicine, University of Malaya, 19 Malaysia. 20 21 b Department of Cardiology, Sarawak Heart Centre, Malaysia. 22 23 c 24 Pharmaceutical Services Division, Ministry of Health, Malaysia. 25 d 26 Department of Pharmacy, Clinical Research Centre, Queen Elizabeth 2 Hospital, Sabah, Malaysia. 27 28 e Department of Pharmacy, Sarawak Heart Centre, Malaysia. 29 30 f Administrative Office, Penang General Hospital, Malaysia. 31 32 g Division of Cardiology, Clinical Research Centre, Queen Elizabeth 2 Hospital, Sabah, Malaysia. 33 http://bmjopen.bmj.com/ 34 h Department of Cardiology, National Heart Institute, Kuala Lumpur, Malaysia 35 36 i 37 Division of Cardiology, Penang General Hospital, Penang, Malaysia. 38 j 39 Division of Cardiology, Department of Medicine, Faculty of Medicine, University of Malaya, 40 Malaysia. 41 on October 1, 2021 by guest. Protected copyright. 42 k Julius Centre University of Malaya, Department of Social and Preventive Medicine, Faculty of 43 44 Medicine, University of Malaya, 50603 Kuala Lumpur, Malaysia 45 46 *Corresponding Author 47 48 Email address: [email protected] Telephone number: + 6012-4281182 49 50 51 Postal address: Department of Social and Preventive Medicine, Faculty of Medicine, University of 52 Malaya, Jalan Universtiti, 50603, Kuala Lumpur, Malaysia. 53 54 55 WORD COUNT 4124 56 57 58 59 1 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 2 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 ABSTRACT 4 5 Objectives: 6 7 8 Limitations in the quality and access of cost data from low- and middle-income countries constrain 9 10 the implementation of economic evaluations. With the increasing prevalence of coronary artery 11 disease in Malaysia, cost information is vital for cardiac service expansion. We aim to calculate the 12 13 hospitalization cost of percutaneous coronary intervention (PCI), using a data collection method 14 customized to local setting of limited data availability. 15 16 For peer review only 17 Design: 18 19 This is a cross-sectional costing study from the perspective of healthcare providers, using top-down 20 21 approach, from January-June 2014. Cost items under each unit of analysis involved in the provision 22 of PCI service were identified, valuated and calculated to produce unit cost estimates. 23 24 25 Setting: 26 27 28 Five public cardiac centres participated. All the centres provide full-fledged cardiology services. 29 They are also the tertiary referral centres of their respective regions. 30 31 32 Participants:

33 http://bmjopen.bmj.com/ 34 The cost was calculated for elective PCI procedure in each centre. PCI conducted for 35 36 urgent/emergent indication or for patients with shock and hemodynamic instability were excluded. 37 38 39 Primary and secondary outcome measures: 40 41 The outcome measures of interest were the unit costs at the 2 units of analysis, namely cardiac ward on October 1, 2021 by guest. Protected copyright. 42 43 admission and cardiac catheterization utilization, which made up the total hospitalization cost. 44 45 Results: 46 47 48 The average hospitalization cost ranged between RM11,471 (USD 3,186) and RM 14,465 (USD 49 4,018). PCI consumables were the dominant cost item at all centres. The centre with daycare 50 51 establishment recorded the lowest admission cost and total hospitalization cost. 52 53 54 Conclusions: 55 56 Comprehensive results from all centres enable comparison at the levels of cost items, unit of analysis 57 58 59 2 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 3 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 and total costs. This generates important information on cost variations between centres, thus 4 providing valuable guidance for service planning. Alternative procurement practices for PCI 5 6 consumables may deliver cost reduction. For countries with limited data availability, costing method 7 8 tailored based on country setting can be used for the purpose of economic evaluations. 9 10 11 KEYWORDS 12 13 14 15 Costing approach, top-down costing, research methods, hospitalization cost, cardiac centres, low- 16 and middle-income Forcountries peer review only 17 18 19 20 Registration: 21 22 23 24 Malaysian MOH Medical Research and Ethics Committee (ID: NMRR-13-1403-18234 IIR) 25 26 27 28 ARTICLE SUMMARY 29 30 31 Strengths and limitations of this study 32

33 http://bmjopen.bmj.com/ 34 35 • A multi-centre costing analysis using standardized collection methods can lead to within- and 36 37 between-centre comparison at multiple levels, from cost items, units of analysis to overall 38 hospitalization cost. 39 40 • The non-participation of private cardiac centres may limit the generalizability of the results. 41 on October 1, 2021 by guest. Protected copyright. 42 • Top-down costing approach applied in this study produced an average estimate cost per 43 patient and enabled an objective comparison of resource consumption and hospitalization 44 45 cost between different centres. 46 47 • However, this average cost estimates are insufficient for in-depth analysis at patient-level or 48 determination of cost predictors via regression analysis. 49 50 • This alternative costing methods we devised to overcome the data limitations in our setting 51 52 can be adapted by like-minded researchers in other low- and middle-income countries. 53 54 55 56 INTRODUCTION 57 58 59 3 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 4 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 Economic evaluation is an important component of healthcare delivery. With the increasing 4 prevalence of diseases and advancement of medical technologies, healthcare costs continue to 5 6 escalate. In high-income countries (HIC) with an easy access to high quality and freely available 7 8 data, there are well-established guidelines for healthcare economic evaluation.[1] Therefore, costing 9 analysis of healthcare services are commonly applied to assess the impact of investments in disease 10 11 prevention and treatment, and to guide decision-making in budget allocation and service planning.[2] 12 13 In contrast, the lack of infrastructure and financial support for evidentiary data capture in low and 14 middle-income countries (LMIC) results in a paucity of easily accessible and reliable cost data. This 15 16 becomes a huge challengeFor to thepeer healthcare providersreview in LMIC towardsonly the uptake of research and 17 application of economic evaluation.[3-6] 18 19 20 21 Cardiovascular disease (CVD) is a leading cause of morbidity and mortality in Malaysia. In 2013, 22 23 CVD accounted for 24.71% of total mortality and was one of the top five causes of 24 hospitalization.[7] CVD encompasses a range of conditions including coronary artery disease 25 26 (CAD), cerebrovascular disease, peripheral artery disease, and heart failure. Among all, CAD 27 28 accounts for the highest prevalence and mortality. Percutaneous coronary intervention (PCI) is a 29 common treatment modality for CAD due to its safety profile in terms of lower mortality and 30 31 complications.[8, 9] In many countries, the availability of coronary catheterization facilities and 32 access to PCI is a key performance indicator of the healthcare service. However, the need for 33 http://bmjopen.bmj.com/ 34 sophisticated laboratory, highly skilled clinical staff, and costly consumables such as cardiac stents 35 36 often drive up the cost of PCI. This can be a constraining factor towards its service establishment and 37 delivery in resource-limited countries. 38 39 40

41 Currently, more than 60 public, private and teaching institutions, most of which are tertiary-level on October 1, 2021 by guest. Protected copyright. 42 referral hospitals located in urban areas, are performing approximately 12000 PCIs annually in 43 44 Malaysia. This service is not available in district hospitals of lower tiers. In the public healthcare 45 system, cardiology services are provided by teaching hospitals and state-level Ministry of Health 46 47 hospitals. In addition, Malaysia has a highly specialized national heart institute, which was 48 49 established as a corporatized entity, and given autonomy in terms of financial and staff autonomy, 50 even though it is fully owned by the government. Cost of medical procedures and hospitalization 51 52 often depend on the type of hospitals. International studies have reported cost differences between 53 54 public- or private-owned hospitals, teaching or non-teaching institutions.[10, 11] However, local 55 evidence are lacking; as the development of health economics is still in its infancy stage in Malaysia, 56 57 58 59 4 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 5 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 and efforts to conduct economic evaluations are hampered by limited access to financial data and 4 non-computerized patient clinical data. Policy makers and healthcare professionals will be interested 5 6 to gain insight into the local cost variation. In order to do so, reliable cost data from cardiac centres 7 8 are needed. 9 10 11 12 OBJECTIVES 13 14 15 With these in mind, we describe a simple, standardized data collection method for a multi-centre 16 For peer review only 17 costing study of cardiac service in Malaysia. Elective PCI is the chosen procedure as it is among the 18 most commonly conducted cardiac procedure with the least heterogeneity among the patients and 19 20 providers. We aim to outline the modifications made to existing guidelines. By presenting the cost 21 22 results from five cardiac centres, we hope to ascertain the cost variation between centres, and to 23 identify potential cost-saving options. 24 25 26 27 METHODS 28 29 30 Study Design 31 32

33 This costing analysis was designed as a hospital-based cross-sectional study, from the perspective of http://bmjopen.bmj.com/ 34 st th 35 healthcare providers. Cost data collection was conducted from January 1 2014 to June 30 2014; 36 using a top-down costing approach. All cost estimates were presented in the local currency, 37 38 Malaysian Ringgit (RM), whereby USD 1=RM 3.60 at the time of study. 39

40 41 Patient Population and Clinical Data Collection on October 1, 2021 by guest. Protected copyright. 42 43 44 PCI can be conducted as an emergency or elective procedure. Patients who undergo emergency PCI 45 46 often have more severe comorbidities and disease presentation, leading to higher resource 47 48 consumption and hospitalization cost. They might develop complications post PCI thus requiring 49 admission to an intensive care unit. On the contrary, patients who undergo elective PCI have lower 50 51 risk profile and likely fewer complications. As a relatively homogenous group, patients undergoing 52 elective PCI would be a better proxy to analyze hospital-level cost. As such, we restricted our 53 54 analysis to elective PCI, excluding patients with urgent/emergent indication for PCI, or with shock 55 56 57 58 59 5 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 6 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 and hemodynamic instability. This definition is compatible with the definition of elective PCI 4 adopted by other studies.[12, 13] 5 6 7 8 Study Sites 9 10 For the cost data collection, 13 public and private cardiac centres were invited to collaborate in this 11 12 study. However, only five public cardiac centres agreed to participate. Being the tertiary referral 13 14 hospitals, all centres are located in the most urbanized area of their respective region. They provide 15 full-fledged cardiology services and is staffed by at least one interventional cardiologist. Centre I is a 16 For peer review only 17 semi-corporatized university teaching hospital while Centre II, III, and IV are government public 18 19 hospitals. Centre V is a specialized heart centre established as a corporatized entity. The range of 20 centres was able to represent possible cost variations stemming from the heterogeneity in hospital 21 22 characteristics, treatment preferences and geographical locations. Table 1 outlines the main 23 description of each centre. 24 25 26 27 28 29 30 31 32

33 http://bmjopen.bmj.com/ 34 35 36 37 38 39 40

41 on October 1, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 6 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 7 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 Table 1 Background Information of Study Centres 4 5 6 Centre I II III IV V 7 8 Ownership Type Semi- Public Public Public Corporatized 9 corporatized 10 Year of Initiation of PCI service 1987 1994 2011 2010 1992 11 12 13 *Interventional Cardiologist 5 3 1 1 15 14 15 *Fellow Cardiologists 12 8 7 0 12 16 For peer review only 17 *Total Hospital Admission 34414 29336 1744 9340 8485 18 19 20 *During the study period of January-June 2014 21 22 23 24 25 26 27 28 29 30 31 32

33 http://bmjopen.bmj.com/ 34 35 36 37 38 39 40

41 on October 1, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 7 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 8 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 Cost Data Collection 4 5 6 Due to the inherent characteristics of financial and medical record keeping in each healthcare system, 7 there is no pre-existing data collection tool suitable for use by all countries. For this study, our team 8 9 of health economists and interventional cardiologists created a data collection tool using MS Excel, a 10 programme commonly used in local clinical and research facilities. Figure 1 outlines the steps of the 11 12 cost data collection. For each centre, a data enumerator and a site coordinator were assigned. The 13 14 data enumerator was in charge of the data collection from the respective departments. Site 15 coordinators were senior personnel with vast knowledge on the daily clinical and financial operations 16 For peer review only 17 of the centre, especially pertaining to cardiology department and PCI procedures. 18 19 20 Pilot study was conducted at Centre IV. Improvements were made based on shortcomings identified 21 22 and feedbacks from data enumerator, site coordinator and data providers. Training workshops were 23 organized to deliver topics such as the purpose of the study, definitions of cost items, introduction of 24 25 the data collection tool and instructions on how to use the tool. Upon completion of data collection 26 27 by data enumerator, site coordinator would check the completeness of the data and conduct face 28 validity check to identify unjustifiable outliers. The compiled cost data was then forwarded to the 29 30 primary investigator and assessed for its consistency and validity. Any unit cost outliers (defined as 31 less than one-tenth or more than 10 times the average unit cost at all centres) were flagged and 32 33 queried with the site coordinators. All the cost data were subsequently reviewed in the presence of http://bmjopen.bmj.com/ 34 35 cardiologists and deemed accurate and justifiable before further calculation and analysis. 36 37 38 Costing Pathway (Figure 2) 39 40

41 Step 1. Identification of unit of analysis on October 1, 2021 by guest. Protected copyright. 42 43 To achieve the most accurate estimate of PCI cost, it is important to identify all the resources 44 45 consumed by the patients. For this study, we outlined the event pathway for the provision of care for 46 47 PCI patients from admission to discharge. Two main units of analysis were identified, namely 48 cardiac ward (CW) and cardiac catheterization laboratory (CL). Based on guidelines in costing 49 50 manual, health economists in our study team listed out various cost items that should ideally be 51 included under each unit of analysis. The final list of cost items depended on general consensus of 52 53 interventional cardiologists as they were the subject-matter experts. Under each unit of analysis, cost 54 55 items that can be directly attributed to patient care are categorized under direct medical cost, while 56 resources that are used by more than one department/unit in the hospital are categorized as overhead 57 58 59 8 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 9 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 cost items. 4 5 6 Step 2. Identification of Cost Items 7 8 9 For CW, direct medical cost items were labor, capital, consumables, and medication; whereas 10 11 overhead costs included utility, dietary, ancillary service and hospital support service. There were 12 less cost items under CL, the second unit of analysis. Medications were prescribed and served at 13 14 CW, thus not taken as a cost item under CL. Dietary and ancillary support service were also provided 15 only once under CW. 16 For peer review only 17 18 19 Step 3. Valuation of Cost Items (Table 2) 20

21 As reference cost is still under development for public hospitals in Malaysia, actual calculation of 22 23 unit cost for all the cost items was conducted using a top-down approach. The subsequent sections 24 25 discussed the valuation of each cost item. 26 27 28 Labour 29 30 This referred to the full salary inclusive of wages and employers’ social contributions paid for the 31 staff. A list of all the clinical staff involved in the provision of PCI service were used to obtain their 32

33 payroll register from the finance department. For clinical staff stationed full time in the same unit, http://bmjopen.bmj.com/ 34 for example nurses and attendants in CW or laboratory technicians and radiographers in CL, their 35 36 salaries were summed up in total as all of their productivity were contributed to their respective 37 38 units. As for doctors, their work scope may spread across several units; including conducting 39 procedures in the CL, treating patients in the specialist clinic, and leading the ward rounds in CW. 40

41 The interventional cardiologists in our study team were also the head of cardiology department at on October 1, 2021 by guest. Protected copyright. 42 43 each centre. Therefore, their expert opinion was sought on the portion of time spent on various 44 activities by doctors of different levels at their centre. The apportioned labour cost of doctors based 45 46 on working hours assumed that all doctors were equally productive in delivering patient-related care 47 and no idle time was spent. 48 49 50 51 Capital 52 This included fixed, one-time expenses incurred on the purchase of land, building, construction, and 53 54 equipment. As the main intention of this study was to obtain the operational cost of PCI procedure, 55 56 land and building costs were excluded. Only equipment such as machine, furniture and medical 57 58 59 9 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 10 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 instrument with cost higher than RM 1000 and economic useful life exceeding 1 year were included. 4 The complete list of asset in CW and CL, together with their price and year of purchase were 5 6 obtained from the inventory list. Using the straight-line depreciation approach, it was assumed that 7 8 the services from the asset were divided equally over the useful life. A discount rate of 3% and a 9 useful life of 5 years were used in conformity with most economic evaluations.[14] The equivalent 10 11 annual cost (EAC) of each asset was calculated by annuitizing the capital outlay using a discount rate 12 13 with their respective useful life years. 14 15 16 Consumables For peer review only 17 This included items that are disposable in nature and require regular replacement, such as syringes, 18 19 cotton swabs and needles. The costs of general consumables were obtained from the procurement 20 21 section of pharmacy department. For each of the cardiac centre, consumable provision may occur at 22 different levels such as individual wards or departments. Depending on the best available cost 23 24 information, workload ratio from the corresponding unit was used as the allocation basis to calculate 25 26 the consumable cost. As for PCI-specific consumables such as cardiac stents, catheters, wires and 27 balloons, the quantity and cost of purchase were retrieved from the purchase ledger. 28 29 30 Medication 31 32 Under the drug purchasing mechanism in public hospitals, general medication and cardiac-specific 33 http://bmjopen.bmj.com/ 34 medication were purchased by central pharmacy under separate budget. General medications are 35 often stored as floor stock in the ward whereas cardiac-specific medications would be distributed to 36 37 cardiology department. Appropriate workload ratio was used as the allocation basis to determine the 38 39 cost of general and cardiac-specific medication. 40 41 on October 1, 2021 by guest. Protected copyright. 42 Utility 43 Utility cost referred to expenses incurred on electricity, water, telephone and Internet service. The 44 45 total bill was obtained from the annual financial account. Hospital floor plan from the engineering 46 47 department was obtained to calculate the space ratio, i.e. the percentage of square metres of physical 48 space occupied by the unit of analysis (CW or CL) in comparison to the hospital indoors area. Space 49 50 ratio was the allocation basis to derive the utility cost of CW and CL. 51

52 53 Hospital Support Service 54 55 For all the participating centres, hospital support services such as housekeeping, laundry, waste 56 management, building, and equipment maintenance were outsourced as an annual contract to a 57 58 59 10 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 11 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 private company. The annual contract cost was obtained from the finance department. Similar to 4 utility cost, space ratio was used as the allocation basis. 5 6 7 8 Dietary 9 Cost of meal preparation by the in-house dietary unit was derived by allocating the total dietary cost 10 11 to CW based on inpatient bed days. 12 13 14 Ancillary Service 15 16 Services from pharmacy,For radiology, peer laboratory, review and rehabilitation only departments are usually provided 17 to all hospital users. Due to time and budget constraints in this study, separate costing analysis was 18 19 not conducted to assign the cost of each of these department to the individual patients. Instead, we 20 21 followed the recommendation by Hendriks et al. and grouped them as a context-specific overhead 22 category, labelled as ancillary service.[15] Its total cost would be the summation of the cost of 23 24 general consumables supplied, the full salary pay-out for all the staff, the utility and hospital support 25 26 services cost calculated in the same manner as mentioned above. 27 28 29 Based on a report in Malaysia, 60% of clinical support services such as those provided by the 30 ancillary departments in our study can be attributed to the inpatient use, compared to 40% for 31 32 outpatient use.[16] Therefore, the same 60:40 proportion was applied in this study for the cost of 33 http://bmjopen.bmj.com/ 34 ancillary services. Using the workload ratio of CW to hospital inpatient bed days as allocation basis, 35 the cost of ancillary service was calculated. 36 37 38 39 Step 4 Calculation of Unit Cost 40 41 Upon valuation, the summation of all cost items gave rise to the total direct medical and overhead on October 1, 2021 by guest. Protected copyright. 42 43 costs incurred at each unit of analysis. To delineate the cost incurred by patients who underwent 44 elective PCI, suitable denominators were applied. For CW, the inpatient bed day was deemed as an 45 46 objective indicator for the workload. By dividing the total cost of CW with the number of CW 47 48 inpatient bed days during the study period, we were able to produce average cost per bed day. The 49 cost per bed day was further multiplied by an average length of stay (ALOS), giving rise to the 50 51 average cost per admission. The ALOS for an elective PCI patient was decided to be 3 days via 52 53 consensus of all the interventional cardiologists in the study. For Centre III, the weighted admission 54 cost was calculated based on the ratio of elective PCI patients admitted to inpatient ward and 55 56 daycare. 57 58 59 11 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 12 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 4 For CL, the second unit of analysis, the number of CL procedures conducted was used as the 5 6 denominator to derive the cost per CL utilization. As PCI consumables were not shared by all 7 8 procedures conducted in CL, we applied a separate denominator i.e. the number of PCI procedures, 9 to calculate the PCI consumables cost per procedure. Together, this gave rise to cost per PCI 10 11 procedure in CL. 12 13 14 The summation of cost per admission and cost of per PCI procedures in CL led to the total 15 16 hospitalization cost Forof a patient peer who underwent review elective PCI. only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32

33 http://bmjopen.bmj.com/ 34 35 36 37 38 39 40

41 on October 1, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 12 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 13 Total Cost of Cost Items Items Cost of Cost Total Allocation Method for Calculation of of Calculation for Method Allocation Total salary of all staffs staffs of all salary Total Apportioned working hours * salary payout of payout salary * hours working Apportioned doctors Allocation of total cost to workload ratio of CW CW of ratio workload to cost total of Allocation Allocation of total cost to CW or CL to or CL to CW cost total of Allocation ratio workload corresponding consumable inpatient Total= cost consumable inpatient total to days bed CW of* ratio cost days bed hospital inpatient bed days. days. bed inpatient hospital cost to workload ratio of CW bed days over over days bed CW of ratio workload to cost Allocation of total inpatient ancillary services services ancillary inpatient total of Allocation inpatient versus outpatient departments. departments. outpatient inpatient versus Assumption of 60:40 ratio of utilization by by of utilization ratio 60:40 of Assumption CW or CL over total indoor area. area. indoor total over or CL CW Allocation of total contract cost to space ratio of to space cost contract total of Allocation Centre I and and II. I Centre Allocation of total cost to workload ratio for ratio workload to cost total of Allocation was available. available. was and cardiac-specific medication supplied to CW CW to supplied medication cardiac-specific and For Centre III, IV and V, actual cost of general of general cost V, actual and IV III, Centre For Total equivalent annual cost (EAC) of all assets assets all costof (EAC) annual equivalent Total procedure per cost Average Allocation of total utility bill to space ratio of to space bill utility total of Allocation area. indoor total over or CL CW bed days over hospital inpatient bed days days bed inpatient hospital over days bed Eg. For Centre I, allocated CW general general CW I, allocated Centre For Eg. No further allocation needed. needed. allocation further No

(Source ofData) (Source Not Applicable Applicable Not Not Applicable Applicable Not

BMJ Open All Inpatient departments departments All Inpatient CW/CL CW/CL department Cardiology All Inpatient departments departments All Inpatient CW/CL CW/CL Cardiology department department Cardiology http://bmjopen.bmj.com/ Resource Output Data Required for Required Data Output Resource Cost Item Calculation Item Cost

ii. iii. i. iii. ii. i. Time spent for different work activities activities work different for spent Time Department) Cardiology of (Head (Patient record office, CW census, CL procedure log) procedure CL census, CW office, record (Patient (Patient record office, CW census) CW office, record (Patient Hospital inpatient bed days days bed inpatient Hospital CW bed days bed CW (Engineering) Space area of hospital indoor area, CW and CL CL and CW area, indoor of hospital area Space (Patient record office, CW census, CL procedure logs) logs) procedure CL census, CW office, record (Patient (based on the best cost data available) dataavailable) cost best the on (based Workload Ratio in terms of inpatient bed days at: days bed inpatient of in terms Ratio Workload

Workload ratio in terms of inpatient bed days at: days bed inpatient of terms in ratio Workload dataavailable) cost best the on (based CW and hospital inpatient bed days days bed inpatient hospital and CW census) CW office, record (Patient Total number of PCI procedures procedures of PCI number Total logs) procedure (CL Space area of hospital indoor area, CW and CL CL and CW area, indoor of hospital area Space (Engineering)

on October 1, 2021 by guest. Protected copyright. For peer review only For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml (Source of Data) (Source (Finance) (Finance) All Inpatient departments departments Inpatient All Cardiology department department Cardiology CW/CL CW/CL All Inpatient departments departments Inpatient All CW/CL CW/CL Cardiology department department Cardiology Cost Items Items Cost

Cost Data Required for Valuation of of for Valuation Required Data Cost

iii. ii. i. iii. ii. i. (Procurement section of pharmacy department) department) pharmacy of section (Procurement (based on the availability of any of the below) below) of the any of availability the on (based (Finance, Human Resource, Pharmacy, Pharmacy, Resource, Human (Finance, and utility costs based on space ratio space on based costs utility and all four departments, allocated hospital support support hospital allocated departments, four all Labour cost for all personnel, consumables cost to to cost consumables forpersonnel, all cost Labour (Finance) (Finance) Annual contract cost cost contract Annual (Procurement section of pharmacy department) department) pharmacy of section (Procurement (based on the availability of any of the below) below) of the any of availability the on (based Total cost of general medication supplied to: supplied medication of general cost Total

(Engineering, Finance) Finance) (Engineering, Asset inventory, year and price of purchase price of purchase and year inventory, Asset (Human Resource, Finance) Resource, (Human (Human Resource, Finance) Resource, (Human Total cost of catheters, stents, balloons, and wires. wires. and balloons, stents, of catheters, cost Total ledger) purchase department (Cardiology Total dietary cost cost dietary Total (Dietary) Monthly utility bill utility Monthly Engineering) Engineering)

PCI-

Others Others than doctors other forstaffs payout Salary

Doctor Doctor grades different of fordoctors payout Salary specific specific General General to: supplied consumable of general cost Total

Medi- cation cation Consumables Labour Utility Capital Capital Dietary Dietary Support Support Hospital Hospital Ancillary Ancillary Cost Items Items Cost Table 2 Data Required and Allocation Method for the Calculation of Cost Items of Items Cost for Calculation the Allocation Method and Required Table 2 Data 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 13 of 28 BMJ Open Page 14 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 RESULTS 4

5 6 Based on the methods described, a range of results can be generated (Table 3). Comparisons 7 8 can be made at the levels of cost item, unit of analysis (CW or CL), up to the total cost. It is 9 important to note that the cost was produced via a top-down approach and it represented the 10 11 average cost estimates for a patient admitted for an elective PCI procedure, assuming average 12 13 length of stay of 3 days, with no complications post PCI that required intensive care. This 14 cost also did not reflect any difference in the type and number of stents used for each patient. 15 16 For peer review only 17 The total hospitalization cost ranged from RM11,471 (USD 3,186) to RM 14,465 (USD 18 19 4,018). The CW admission cost accounted for one-fifth or less of the total hospitalization cost 20 21 at all centres. On the contrary, cost incurred at CL was four times that of the admission cost. 22 PCI consumables contributed to the biggest proportion among all the cost items, accounting 23 24 for more than half of the total costs. 25 26 27 Total hospitalization cost was the highest at Centre II. Closer scrutiny revealed that average 28 29 cost of PCI consumables of Centre II, and its proportion over the total hospitalization cost 30 were highest across the 5 centres. In contrast, Centre III, the centre with a daycare service for 31 32 PCI patients, recorded the lowest cost of CW admission. This partly contributed to cheapest 33 http://bmjopen.bmj.com/ 34 hospitalization cost in Centre III, at 12.3% to 26.1% lower compared to the other centres. 35 Direct medical cost items were generally more expensive than overhead cost items in both 36 37 CW and CL. For CW admission, medication and labour costs were the highest. As for CL 38 39 utilization, capital cost ranked the highest at Centre I and V, the two corporatized hospitals. 40

41 on October 1, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 14 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 15 1558 3682 2709 595.19 63,584.90 12,697.27 109,094.76 112,959.66 222,997.80 154,201.25 118,721.94 159,664.13 126,005.76 2,530,478.96 3,619,363.29 5,130,845.80 2,191,477.81 1,333,604.78 12,743,046.50 12,743,046.50 11,502,742.46 8,179.11 (57.6%) 8,179.11 4,246.12 (29.9%) 4,246.12 1,785.56 (12.6%) (12.6%) 1,785.56 14,210.79 (100%) 14,210.79 12,425.23 (87.4%) 12,425.23

251 940 2526 830.18 830.18 15,417.61 70,137.06 86,215.03 19,857.02 11,603.16 26,269.15 52,784.55 80,074.16 500,037.88 782,479.78 334,733.31 575,258.55 1,702,805.64 2,097,039.64 1,244,978.02 2,152,343.50 2,152,343.50 Centre IV IV Centre V Centre 8,575.07 (66.6%) 8,575.07 1,811.50 (14.1%) 1,811.50 2,490.55 (19.3%) 2,490.55 12,877.11 (100%) 12,877.11 10,386.57 (80.7%) 10,386.57

282 1485 2887 479.59 479.59 49,302.66 87,785.37 47,716.53 62,189.01 62,473.17 43,305.00 Daycare Daycare 778,561.21 687,789.25 398,523.57 345,683.73 250,095.62 111,235.99 461,883.47 2,001,962.06 1,384,582.52 2,379,464.50 2,379,464.50

9,785.95 (85.3%) 9,785.95 8,437.82 (73.6%) 8,437.82 1,348.13 (11.8%) (11.8%) 1,348.13 11,471.04 (100%) 11,471.04 Centre III Centre 3892 818.53 1,685.09 (14.7%) (14.7%) 1,685.09 64,321.29 58,374.60 193,787.82 558,577.96 124,948.95 222,476.63 622,613.16 Inpatient Inpatient 3,185,726.21 1,340,625.80

257 1118 2786 647.87 22,559.11 34,634.59 94,791.64 97,277.64 95,143.27 35,375.00 880,833.45 614,370.96 359,953.93 874,136.64 214,708.34 287,242.14 106,302.47 BMJ Open 1,912,352.04 1,804,977.14

2,778,554.00 2,778,554.00 Centre II Centre http://bmjopen.bmj.com/ 1,710.51 (11.8%) 1,710.51 1,943.62 (13.4%) 1,943.62 14,465.64 (100%) 14,465.64 12,522.00 (86.6%) 12,522.00 10,811.49 (74.7%) 10,811.49

580 1084 3129 906.68 906.68 83,040.43 87,132.83 26,075.00 420,404.41 201,404.15 116,035.54 109,096.89 878,721.52 114,473.40 666,064.92 382,311.39 1,055,939.18 2,839,335.96 4,143,031.92 2,837,003.70 4,504,850.00 4,504,850.00 7,766.98 (54.3%) 7,766.98 3,821.99 (26.7%) 3,821.99 2,720.04 (19.0%) 2,720.04 on October 1, 2021 by guest. Protected copyright. 14,309.01 (100%) 14,309.01 11,588.97 (81.0%) 11,588.97

For peer review only For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

Total Cost of PCI Consumables Consumables of PCI Cost Total Total Cost of CL of CL Cost Total Labour Labour Capital Consumable Utility Support Hospital Average Cost per Bed Day Bed per Cost Average Total cost of CW ofCW cost Total Labour Labour Capital Consumable Medication Utility Dietary Support Hospital support Ancillary Number of PCI Procedures Procedures PCI of Number Number of CL Procedures Procedures CL of Number Number of CW Bed Days Days Bed CW of Number Medical head Medical

Direct Direct Over- Overhead Overhead Direct Direct

*Weighted admission cost was calculated for Centre III based on the ratio of elective PCI patients admitted to inpatient ward versus daycare. daycare. versus ward inpatient to admitted patients PCI elective of ratio on the IIIbased Centre for calculated was cost admission *Weighted Total Hospitalization Cost, A+B+C A+B+C Cost, Hospitalization Total Cardiac Catheterization Laboratory Utilization (CL) (CL) Utilization Laboratory Catheterization Cardiac Cardiac Ward Admission (CW) (CW) Admission Ward Cardiac Table 3 Cost Items of Each Unit of Analysis Contributing to Total Hospitalization Cost of Elective PCI at All Centres All Centres PCI Cost at of Elective to Hospitalization Total Contributing of Analysis Each Unit of Items Table 3 Cost Ringgit) (Malaysian Item ofCost Cost Total I Centre Average Cost of PCI procedure in CL, B+C B+C in CL, procedure ofPCI Cost Average Average Cost of PCI Consumables, C Consumables, ofPCI Cost Average Average Cost per CL utilization, B B utilization, CL per Cost Average Average Cost per Admission, A A Admission, per Cost Average 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 15 of 28 BMJ Open Page 16 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 DISCUSSION 4

5 6 As CVD represents the leading cause of mortality and morbidity in LMIC, expansion of the 7 8 cardiology service is inevitable. However, most LMIC lack guidance for budget and resource 9 planning as they do not have any official economic evaluation guidelines, and the international 10 11 guidelines are not a good fit for LMIC. There is also a lack of infrastructure support to sustain a 12 13 comprehensive health informatics data. These are indispensable components to produce reliable cost- 14 related outcome evidence towards better policy-making. In Malaysia, there is a lack of easily 15 16 accessible informationFor such as peer accurate unit reviewcosts, detailed financial only account and complete patient 17 records, all of which are vital to conduct economic evaluation. In this paper, we outlined the study 18 19 design of a multi-centre costing analysis for elective PCI, based on the limitations in our local 20 21 setting, before presenting the cost output. Overall, the results demonstrated substantial costs 22 associated with the provision of PCI, with PCI consumables being the dominant cost item across all 23 24 centres. 25 26 27 In the course of designing the costing analysis, we made some modifications to previously published 28 29 guidelines. International guidelines recommended the design of a standardized cost collection tool 30 that is universally available and easy to use for all involved parties.[17, 18] Thus, we chose MS 31 32 Excel over other more sophisticated costing software, in view of its wide accessibility and user- 33 http://bmjopen.bmj.com/ 34 friendliness. Many HIC used Diagnosis Related Group (DRG), a system which classifies acute 35 inpatient episodes based on the main clinical condition, as unit of analysis for illness-specific 36 37 costing.[19] However, this is not a common practice in our local hospitals. To overcome this, we 38 39 used an event pathway from admission to discharge to identify cost items under the respective units 40 of analysis. Furthermore, reference cost from national database of HIC are available for cost item 41 on October 1, 2021 by guest. Protected copyright. 42 valuation, such as in the United Kingdom and Australia.[20, 21] In the absence of reference cost, 43 certain literature proposed the application of cost-to-charge ratio, which is a relatively simpler and 44 45 less time consuming method to reflect the actual cost.[22] However, it would be a suboptimal cost 46 47 estimate as public healthcare is highly subsidized in Malaysia and the charges would be a poor proxy 48 for the actual cost incurred. As a result, we conducted actual calculation of unit cost with top-down 49 50 approach. Between top-down and bottom-up approaches, there is no consensus as to which is the 51 gold standard. Typically, top-down approach is more straightforward but the trade-off is a lack of 52 53 sensitivity, especially for cost items that originated from a less homogenous production. On the 54 55 contrary, bottom-up approach is able to produce a more accurate cost estimate, but the main 56 limitation lies within the complexity of its implementation, especially across different study 57 58 59 16 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 17 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 centres.[23] Several authors advocated bottom-up approach but acknowledged that such data may 4 not be easily available in LMIC.[3, 24, 25] As this costing study involved five centres, a full top- 5 6 down approach was applied to standardize the cost data collection in order to produce reliable cost 7 8 estimates for comparison between the centres. 9 10 11 The application of economic evaluation in routine clinical practice is not widespread in Malaysia. 12 13 The lack of familiarity among the staff in healthcare facilities towards the conduct of costing analysis 14 was a major barrier in our study. All private hospitals cited confidentiality issue of financial data 15 16 sharing as reason ofFor non-participation. peer During review data collection, there only were circumstances of 17 bureaucratic resistance towards parting with cost data of sensitive nature. Misperception of the 18 19 costing exercise as a performance audit was another reason behind some of the staff reluctance to 20 21 cooperate. As a result, the relationship between the enumerators and other hospital staff proved to be 22 a vital factor towards a successful data collection. In several instances, site coordinators had to step 23 24 in and used their discretion to ensure that complete data was provided. This highlighted the 25 26 importance of ensuring access of transparent and quality cost data towards the successful 27 implementation of a costing analysis. 28 29 30 There were several limitations. Selection bias existed, as there was no input from private for-profit 31 32 healthcare centres. Despite this, the cost output from this study would be of tremendous value as the 33 http://bmjopen.bmj.com/ 34 major bulk of care burden in Malaysia fell on the public healthcare system. This is reflected in a 35 national survey that showed 65.0% of the Malaysian population sought treatment at public healthcare 36 37 facilities and 52.4% of the total health expenditure was incurred in the public sector.[26, 27] The 38 39 exclusion of land and building costs may reduce the applicability of capital cost findings to rural, 40 district hospitals. However, this has less impact towards costing of highly specialized medical 41 on October 1, 2021 by guest. Protected copyright. 42 procedures such as PCI, as the provision of such services in Malaysia will likely remain in tertiary- 43 level hospitals in urban locations. Another limitation being the potential bias arising from inherent 44 45 differences in the financial record keeping between the centres. This may not be completely 46 47 eliminated despite our best effort to standardize the data collection process, as shown in the 48 derivation of consumable and medication costs. Despite the shortcomings, the key strengths of this 49 50 study lie in the practical applications of its methodological contribution and cost findings. We shared 51 the alternative pathways we devised to overcome barriers in conducting a costing analysis in a LMIC 52 53 setting. To the best of our knowledge, this is the first multi-centre costing analysis of cardiac service 54 55 in Malaysia. Average cost estimates derived from five cardiac centres with different ownership and 56 57 58 59 17 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 18 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 administration characteristics enabled us to compare the resource utilization and cost output at 4 various levels of the costing pathway. 5 6 7 8 While direct cost comparison with other published literatures would not be completely feasible 9 owing to the heterogeneous study populations, different perspectives used in cost estimation, varying 10 11 cost calculation methods and vastly different healthcare systems, our finding of PCI consumables 12 13 being the dominant cost item strikes a similar chord with many studies.[28-30] This shows the 14 importance of an efficient consumable purchasing mechanism, especially for cardiac stents. Policy 15 16 makers should reviewFor the current peer procurement review practice of PCI consumablesonly by individual cardiac 17 centres. Vigorous negotiation via competitive bidding processes, ideally handled via a central 18 19 purchasing agency, may lead to potential cost reduction. Furthermore, results from one of the study 20 21 centres showed that daycare establishment can be an attractive cost-saving strategy. The general 22 consensus among relevant studies found that low-risk patients were good candidates for same-day 23 24 discharge after uncomplicated PCI.[31, 32] Nevertheless, further research in more Malaysian 25 26 hospitals are warranted to study the feasibility of establishing large-scale daycare service for cardiac 27 patients. 28 29 30 Overall, the corporatization and teaching status of hospitals in this study appear to have minor 31 32 influence to the eventual cost. However, the cost differences may also be due to the casemix and 33 http://bmjopen.bmj.com/ 34 outcome of patients at each centre. Further casemix-adjusted analysis could be conducted to analyze 35 the factors that contribute to the difference in the cost output. In addition, results from this study can 36 37 be incorporated with clinical data to facilitate individual-level patient analysis, such as the 38 39 identification of cost predictors via regression analysis. The unit cost derived from the various levels 40 of the care pathway can also be applied to economic evaluation such as cost effectiveness analysis or 41 on October 1, 2021 by guest. Protected copyright. 42 economic modeling in relevant research. All these will provide valuable information to the cardiac 43 centres and policy makers for proper resource and budget allocation. Lastly, while in-hospital 44 45 treatment may account for a substantial portion of overall cost of CAD care, pre- and post-PCI cost 46 47 to the healthcare providers, patients and society may represent hidden economic burden. This is an 48 important area for future research. 49 50 51 52 53 54 55 56 CONCLUSION 57 58 59 18 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 19 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 4 The CVD epidemic in LMIC calls for targeted interventions from all aspects of healthcare delivery. 5 6 Currently, the application of evidence from economic evaluation for health policy decision-making 7 8 in Malaysia is very limited due to the obstacles in conducting costing studies. We devised alternative 9 costing pathways to overcome the barriers and conducted a multi-centre costing analysis for elective 10 11 PCI. The findings highlighted the need for effective procurement practice in view of the high cost of 12 13 PCI consumables. Shorter hospitalization stay via daycare establishment represents another potential 14 cost-saving mechanism. Future studies can be built on our efforts for other medical procedures or 15 16 healthcare service. RecommendationsFor peer from thisreview study would also only be useful for like-minded 17 researchers who wish to conduct similar costing studies in LMIC. 18 19 20 21 22 FOOTNOTES 23 24 Acknowledgements 25 26 27 The authors thank all cardiologists and nurses at each participating centres for their effort in data 28 collection. 29 30 31 Contributors 32

33 http://bmjopen.bmj.com/ OI, MD, LHB, OTK, RMA, WAWA, LSY, LEV and KYL conceived the overall study. All authors 34 35 are involved in the design of the study. WAWA, OI, LHB, OTK and RMA led the implementation of 36 37 the study at each centre; while KYL, LEV, LSY, LA and SH coordinated the data collection. KYL 38 drafted the manuscript. TKO and MD revised manuscript critically for intellectual content. All 39 40 authors read and approved the final manuscript.

41 on October 1, 2021 by guest. Protected copyright. 42 43 Funding 44 45 This work was supported by the Malaysia Ministry of Higher Education High Impact Research MoE 46 47 Grant UM.C/625/1/HIR/MoE/ E000010-20001 and Malaysia Ministry of Health MRG grant (MRG- 48 MOH-2014-02). 49 50 51 52 53 Competing interests 54 55 56 All authors declared no competing interest. 57 58 59 19 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 20 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 Ethics approval 4 5 This study is approved by the Malaysian Ministry of Health Medical Research and Ethics Committee 6 7 (ID: NMRR-13-1403-18234 IIR). 8 9 10 Provenance and peer review 11 12 Not commissioned; externally peer reviewed 13 14 15 Data sharing statement 16 For peer review only 17 No additional data are available 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32

33 http://bmjopen.bmj.com/ 34 35 36 37 38 39 40

41 on October 1, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 20 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 21 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 REFERENCES 4 5 1 Drummond M, Sculpher M, Torrance G, et al. Methods For The Economic Evaluation Of Health 6 Care Programmes. 3rd ed. Oxford: Oxford University Press, 2005. 7 2 Lipscomb J, Yabroff KR, Brown ML, et al. Health care costing: data, methods, current 8 applications. Med Care 2009;47:S1-6 doi: 10.1097/MLR.0b013e3181a7e401 [published Online 9 First: 2009/06/19] 10 3 Mogyorosy Z, Smith PC. The main methodological issues in costing health care services: A 11 literature review. Centre for Health Economics, University of York, 2005. 12 http://www.york.ac.uk/media/che/documents/papers/researchpapers/rp7_Methodological_i 13 ssues_in_costing_health_care_services.pdf. 14 4 von Both C, Jahn A, Fleba S. Costing maternal health services in South Tanzania: a case study 15 from Mtwara Urban District. The European journal of health economics : HEPAC : health 16 For peer review only 17 economics in prevention and care 2008;9:103-15 doi: 10.1007/s10198-007-0048-3 [published 18 Online First: 2007/04/25] 19 5 Sarowar MG, Medin E, Gazi R, et al. Calculation of costs of pregnancy- and puerperium-related 20 care: experience from a hospital in a low-income country. Journal of health, population, and 21 nutrition 2010;28:264-72 [published Online First: 2010/07/20] 22 6 Briggs A, Nugent R. Editorial. Health economics 2016;25 Suppl 1:6-8 doi: 10.1002/hec.3321 23 [published Online First: 2016/01/26] 24 7 Planning and Development Division HIC. Health Facts 2013. Malaysia, 2014. 25 http://www.moh.gov.my/images/gallery/publications/HEALTH FACTS 2014.pdf. 26 27 8 Keeley EC, Boura JA, Grines CL. Primary angioplasty versus intravenous thrombolytic therapy 28 for acute myocardial infarction: a quantitative review of 23 randomised trials. Lancet 29 2003;361:13-20 doi: 10.1016/s0140-6736(03)12113-7 [published Online First: 2003/01/09] 30 9 Van de Werf F. Drug-eluting stents in acute myocardial infarction. The New England journal of 31 medicine 2006;355:1169-70 doi: 10.1056/NEJMe068165 [published Online First: 32 2006/09/15] 33 10 Evans E, Imanaka Y, Sekimoto M, et al. Risk adjusted resource utilization for AMI patients http://bmjopen.bmj.com/ 34 treated in Japanese hospitals. Health Econ 2007;16:347-59 doi: 10.1002/hec.1177 [published 35 Online First: 2006/10/13] 36 11 Polanczyk CA, Lane A, Coburn M, et al. Hospital outcomes in major teaching, minor teaching, 37 38 and nonteaching hospitals in New York state. Am J Med 2002;112:255-61 doi: 10.1016/S0002- 39 9343(01)01112-3 [published Online First: 2002/03/15] 40 12 Negassa A, Monrad ES. Prediction of Length of Stay Following Elective Percutaneous

41 Coronary Intervention. ISRN Surgery 2011;2011:714935 doi: 10.5402/2011/714935 on October 1, 2021 by guest. Protected copyright. 42 13 Shaw RE, Anderson HV, Brindis RG, et al. Development of a risk adjustment mortality model 43 using the American College of Cardiology-National Cardiovascular Data Registry (ACC-NCDR) 44 experience: 1998-2000. Journal of the American College of Cardiology 2002;39:1104-12 45 [published Online First: 2002/03/30] 46 14 Creese A, Parker D. Cost analysis in primary health care: a training manual for programme 47 48 managers. Geneva: World Health Organization, 2000. 49 15 Hendriks ME, Kundu P, Boers AC, et al. Step-by-step guideline for disease-specific costing 50 studies in low- and middle-income countries: a mixed methodology. Global health action 51 2014;7:23573 doi: 10.3402/gha.v7.23573 [published Online First: 2014/04/02] 52 16 Medical Development Division. Guideline on Cost Data Collection for the Casemix System, 53 Malaysian DRG. 2013. 54 17 Batura N, Pulkki-Brannstrom AM, Agrawal P, et al. Collecting and analysing cost data for 55 complex public health trials: reflections on practice. Global health action 2014;7:23257 doi: 56 10.3402/gha.v7.23257 [published Online First: 2014/02/26] 57 58 59 21 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 22 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 18 Drummond M, Brandt A, Luce B, et al. Standardizing methodologies for economic evaluation 4 in health care. Practice, problems, and potential. International journal of technology assessment 5 in health care 1993;9:26-36 [published Online First: 1993/01/01] 6 19 Tan SS, Serde ́n L, van Ineveld BM, et al. DRGs and cost accounting: which is driving which? 7 Diagnosis-related groups in Europe: towards efficiency and quality. Berlin: Open University 8 Press, 2011. 9 20 NHS reference costs 2014 to 2015. United Kingdom, 2015. 10 11 https://www.gov.uk/government/publications/nhs-reference-costs-2014-to-2015. 12 21 Pharmaceutical Benefits Advisory Committee. Manual of resource items and their 13 associated costs. Canberra, Australia, 2009. 14 http://www.health.gov.au/internet/main/publishing.nsf/Content/health-pbs-general-pubs- 15 Manual-content.htm-copy2 16 22 Shwartz M, YoungFor DW, Siegrist peer R. The ratioreview of costs to charges: only how good a basis for 17 estimating costs? Inquiry : a journal of medical care organization, provision and financing 18 1995;32:476-81 [published Online First: 1995/01/01] 19 23 Emmett D, Forget R. The utilization of activity-based cost accounting in hospitals. Journal of 20 21 hospital marketing & public relations 2005;15:79-89 doi: 10.1300/J375v15n02_06 [published 22 Online First: 2005/10/06] 23 24 Tan SS, Rutten FF, van Ineveld BM, et al. Comparing methodologies for the cost estimation 24 of hospital services. The European journal of health economics : HEPAC : health economics in 25 prevention and care 2009;10:39-45 doi: 10.1007/s10198-008-0101-x [published Online First: 26 2008/03/15] 27 25 Clement Nee Shrive FM, Ghali WA, Donaldson C, et al. The impact of using different costing 28 methods on the results of an economic evaluation of cardiac care: microcosting vs gross- 29 costing approaches. Health economics 2009;18:377-88 doi: 10.1002/hec.1363 [published 30 Online First: 2008/07/11] 31 32 26 Malaysia National Health Accounts Unit. Malaysia National Health Accounts Health

33 Expediture Report 1997-2012. Malaysia, 2014. http://bmjopen.bmj.com/ 34 27 Institute for Public Health. The Third National Health and Morbidity Survey (NHMS III) 35 2006. Kuala Lumpur, 2008. 36 28 Varani E, Balducelli M, Vecchi G, et al. Comparison of multiple drug-eluting stent 37 percutaneous coronary intervention and surgical revascularization in patients with multivessel 38 coronary artery disease: one-year clinical results and total treatment costs. J Invasive Cardiol 39 2007;19:469-75 [published Online First: 2007/11/08] 40 29 Manari A, Costa E, Scivales A, et al. Economic appraisal of the angioplasty procedures 41 on October 1, 2021 by guest. Protected copyright. performed in 2004 in a high-volume diagnostic and interventional cardiology unit. J Cardiovasc 42 43 Med (Hagerstown) 2007;8:792-8 doi: 10.2459/JCM.0b013e328012c1b6 [published Online 44 First: 2007/09/22] 45 30 Chino M, Sakamoto M, Sasaki T, et al. [What aspects are inefficient concerning percutaneous 46 coronary intervention in Japan?: International cost-effectiveness comparisons]. J Cardiol 47 2004;44:85-92 [published Online First: 2004/10/27] 48 31 Abdelaal E, Rao SV, Gilchrist IC, et al. Same-day discharge compared with overnight 49 hospitalization after uncomplicated percutaneous coronary intervention: a systematic review 50 and meta-analysis. JACC Cardiovasc Interv 2013;6:99-112 doi: 10.1016/j.jcin.2012.10.008 51 [published Online First: 2013/01/29] 52 32 Brayton KM, Patel VG, Stave C, et al. Same-Day Discharge After Percutaneous Coronary 53 54 InterventionA Meta-Analysis. Journal of the American College of Cardiology 2013;62:275-85 55 doi: 10.1016/j.jacc.2013.03.051 56 57 58 59 22 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 23 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 4 Figure 1 Pathway for the Costing Analysis of Elective PCI 5 6 7 Figure 2 Flow Chart of Task and Person-In-Charge Involved at Each Step of Data Collection Process 8 9 10 11 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32

33 http://bmjopen.bmj.com/ 34 35 36 37 38 39 40

41 on October 1, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 23 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 24 of 28

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 Figure 1 Pathway for the Costing Analysis of Elective PCI 30 31 32 137x95mm (300 x 300 DPI) 33 http://bmjopen.bmj.com/ 34 35 36 37 38 39 40

41 on October 1, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 25 of 28 BMJ Open

1 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32

33 http://bmjopen.bmj.com/ 34 35 36 37 38 39 40

41 on October 1, 2021 by guest. Protected copyright. 42 43 44 45 Figure 2 Flow Chart of Task and Person-In-Charge Involved at Each Step of Data Collection Process 46 47 48 296x419mm (300 x 300 DPI) 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from Page 26 of 28 8-10 8-10 NA 8 8 2 2

NA 12 12 NA 8 8 5-7 5-7 5 5

5 5 4 4

NA NA NA Reported on page # Reported on page 2 2 5 5

BMJ Open http://bmjopen.bmj.com/ on October 1, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

For peer review only ) Indicate study’s) the designcommonlywith used a in term ortitlethe abstract the ) ) theGive eligibility criteria, and sourcesthe methodsand of of selection participants ) ) statisticalDescribe all methods, thoseincluding confoundingusedforcontrol to ) ) Describe any methods used subgroups examine to interactionsand ) If ) applicable, describe methodsanalytical taking account of sampling strategy ) ) in theProvide abstract informativean and balanced summary of what was donewhat wasand found ) Describe) sensitivity any analyses ) how Explain missing data were addressed a a a b e d c b For eachof interest, For variable sourcesdatagive of and details of methods of assessment (measurement). Describe comparability of of comparability assessment methods isthanthere more if one group ( ( why why

applicable collection collection ( ( ( ( Recommendation cross-sectional studies of cross-sectional reports in be included should that items of Statement—Checklist (v4) 2007 STROBE

6 1 ( 3 specific objectives, State including prespecified any hypotheses 5 the setting, Describe locations, relevantand dates, including periods exposure,of recruitment, follow-up, data and 9 any Describe efforts to address sourcespotential of bias 7 define Clearly outcomes,all exposures,predictors, potential confounders,effect and modifiers. diagnostic Give criteria, if 4 key Present of studyelements in thedesign early paper 8* 10 10 how Explain studythe size was arrived at # Item Item Results

Statistical Statistical methods 12 ( Quantitative variables 11 how Explain quantitative were handledvariables in analyses.the If applicable, describe which chosengroupings were and Study size Bias Data sources/Data measurement Variables Participants Setting Study design Methods Objectives Background/rationale 2 thebackground Explain scientific for rationale investigation the and being reported Introduction Title andTitle abstract Section/Topic

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from 17-18 17-18 NA NA

18 18 17 17 16 16

15 15 NA NA NA NA 15 (only unadjusted 15 are estimates here)relevant 19 19 NA NA

BMJ Open http://bmjopen.bmj.com/ on October 1, 2021 by guest. Protected copyright. For peer review only For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml ) Report) boundaries category when continuous variables were categorized ) ) unadjustedGive estimates if applicable, and, confounder-adjusted estimates their and precision95% (eg, confidence ) If ) relevant, If consider translating estimatesrelativerisk of into absolute forrisk meaningfulperioda time c b a similar studies,similar and other relevant evidence confounders magnitude of magnitude potential biasany confirmed eligible,confirmed included in study, the completing follow-up, analysed and interval). Make interval). clear confounderswhich were adjusted they why and for were included which the presentthe which is based article

( ( number Indicate (b) of participants missingwith eachfor data of interest variable Consider (c) diagram use flow of a reasons Give (b) non-participationfor each at stage 18 18 results Summarise key to reference withstudy objectives 16 16 ( 22 22 sourcethe Give of funding and role the of fundersthe thefor studypresent if and, applicable, originalthe for study on 19 19 Discuss of study, the takinglimitations into sourcesaccount biasof potential imprecision.or Discussboth direction and 13* 13* Report (a) numbers of individuals each at stage of study—eg numbers eligible, potentially examinedfor eligibility, An Explanation and Elaboration article discussesitem each checklist and gives backgroundmethodological and published examples of transparent reporting. The STROBE checklist is best used in this(freelywith conjunction available onarticle sites the Medicine of PLoS Web http://www.plosmedicine.org/,at Annals of Internal Medicine at http://www.annals.org/, http://www.epidem.com/).Epidemiology at and Information on InitiativeSTROBE the is www.strobe-statement.org. available at information*Give separately cases for and controls studiesin forcase-control applicable, if and, exposedunexposed and groups in cohort and studies.cross-sectional Note: Funding Other information Generalisability 21 Discuss(external generalisabilitythe of studythe validity) results Interpretation 20 cautious Give interpretationa of resultsoverall limitations,considering objectives, multiplicity of results analyses, from Limitations Key Key results Discussion Other analyses 17 other Report analyses done—eg analyses of subgroups sensitivity interactions, and and analyses

Main resultsMain Outcome Outcome data 15* numbers Report of outcome events or summary measures

Descriptive data 14* characteristicsGive (a) participantsof study (eg demographic, social) clinical, and information on exposures potential and

Participants 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 27 of 28 BMJ Open: first published as 10.1136/bmjopen-2016-014307 on 28 May 2017. Downloaded from Page 28 of

BMJ Open http://bmjopen.bmj.com/ on October 1, 2021 by guest. Protected copyright. For peer review only For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60