Barriers and facilitators to hospital pharmacists’ engagement in medication safety activities: a qualitative study using the theoretical domains framework


Mohd Taufek et al. (2022) Journal of Pharmacy, 2(1), 31-35 

 

  

Page 31  

 

 

 

Validity of claims database compared with the 
electronic medical record of private health clinics 
in Malaysia: A pilot study 

Nor Hidayah Mohd Taufek1, Norny Syafinaz Ab Rahman1, Nor Ilyani Mohamed Nazar1, Siti Halimah 
Bux1, Che Suraya Zin1* 

ABSTRACT 

Introduction: The validity of health insurance claims data in private health sectors has been widely 
reported in many developed countries to monitor details of private healthcare utilisation. Little is 
known regarding the data validity of private health care services and insurance claims in Malaysia. 
This pilot study aimed to validate the claims data from a private health insurance database, using 
electronic medical records (EMR) at the private clinics as the gold standard. 
 
Method: Patients’ data were retrieved from the PMCare health insurance database from 2016-2019 
recorded for International Islamic University Malaysia employees. Patients were sampled from the 
PMCare database and manually compared with data from EMR of selected private panel clinics. 
Data were analysed for descriptive statistics using Microsoft Excel 2013. 
 
Results: A total of four panel clinics consented to the study, and data were available for 2016, 2017 
and 2019. The number of observations obtained from 118 patients (male = 63, female = 55) was 
386, with the most common diagnosis reported in the PMCare database was acute upper 
respiratory tract infection (63.6%). Total accuracy between PMCare and EMR data was 91.5%, with 
an 8.5% difference or inaccuracy. Percentage accuracy was varied between different clinics (A= 
92.6%, B=84.7%, C=98.6%, D=82.6%). 
 
Conclusion: Data submitted to PMCare claims by private health clinics had high accuracy (>90%) 
and is acceptable for research and other applications. Future studies should investigate the 
differences in clinic-based practice for documenting the identified types of discrepancies to improve 
the accuracy of private health insurance databases. 

ARTICLE HISTORY:  

Received: 20 August 2021 
Accepted: 31 January 2022 
Published: 31 January 2022 

 

KEYWORDS:  

Validity, claim database, 
electronic medical record, private 
health clinics, accuracy. 

 

HOW TO CITE THIS 
ARTICLE: 

Mohd Taufek, N. H., Ab Rahman, 
N. S., Mohamed Nazar, N. I. Bux, 
S. H. & Zin C. S. (2022). Validity of 
claims database compared with 
the electronic medical record of 
private health clinics in Malaysia: A 
pilot study. Journal of Pharmacy, 
2(1), 31-35. 

 

doi: 10.31436/jop.v2i1.107 

 

Authors’ Affiliation: 

1 Big Data Research Group, Department of Pharmacy Practice, Kulliyyah of Pharmacy, 

International Islamic University Malaysia, 25200 Kuantan, Pahang, Malaysia 

*Corresponding author: 

Email address: chesuraya@iium.edu.my 



Mohd Taufek et al. (2022) Journal of Pharmacy, 2(1), 31-35 

 

  

Page 32  

Introduction 

 
The healthcare system in Malaysia has been primarily 

divided into private and public sectors. High 

concentrations of private practices utilised by the 

community are covered by private health insurance 

companies (Thomas et al., 2011). Many private practices 

have also been affiliated as panel health clinics to 

government and private organisations which subsidise 

certain medical insurance coverage as health benefits for 

their employees with variable premiums charges (Chua 

&Cheah, 2012). Private health insurance companies record 

the data on patients’ claims such as diagnosis, prescribed 

medications and cost submitted by the panel health clinics 

from their medical records. The format for the claims 

database could be different between companies, but it 

generally collects information on prescriptions, procedures 

and administrative costs of inpatients and outpatients.  

Claims data are useful to monitor details regarding 

healthcare utilisation, including disease and prescription 

patterns. However, it requires high accuracy with good 

validity and reliability components to be tested so that the 

characteristics and limitations of the dataset are well 

understood (Iwamoto et al., 2015; Du et al., 2006). 

Validity studies are crucial owing to increased reliance on 

health insurance claims, and thus the credibility of such 

data should be confirmed (Koram et al., 2019). There have 

been limited validation studies reported in the Asia Pacific 

region, including Australia, South Korea, New Zealand, 

Thailand, Singapore and Japan (Koram et al., 2019). To 

our knowledge, there have been no studies available 

reporting the validity of the claims database in Malaysia. 

Little is known whether prescription patterns, medication 

utilisation, and cost can be accurately obtained from 

private insurance databases. It is important to test the 

accuracy of insurance database documentation to ensure 

the credibility of research utilising databases. Validity of 

database of private insurance companies will facilitate 

research towards sustainable reimbursement policy, 

optimal utilisation of medications and patient outcomes. 

A review has highlighted that the gold standard for 

database validity research was obtained from medical 

records, registry data, self-reported questionnaires and 

other data sources (Koram et al., 2019). Validated 

outcomes included variables such as medical conditions 

(Takeda et al., 2016), risk monitoring (Tomlin, Reith & 

Woods, 2017) and disease‐specific comorbidities 

(Yamana et al., 2017). The present study aimed to 

investigate the accuracy of the claims data of a private 

health insurance company, PMCare, compared with 

electronic medical records (EMR) of private panel clinics 

that are covered by PMCare. 

 

 

 

 

Methodology 

 
This study was approved by the International Islamic 

University Malaysia Ethical Committee (IREC-2019-

212). It was conducted from February to July 2020 at panel 

private health clinics in two states of Pahang and Selangor, 

Malaysia. The results were reported in an aggregated 

manner using de-identified data. Informed consent was not 

required as this study did not involve direct patient 

interaction. 

This retrospective pilot study collected data for 2016-

2019 provided by the private health insurance company 

PMCare for outpatient settings.  The PMCare insurance 

was subsidised by International Islamic University 

Malaysia (IIUM) for its employees. Based on the policy, 

the claim limit was set at Ringgit Malaysia (RM) 45 per 

visit per patient. The panel private health clinics (PHC) 

were identified to collect the original data from EMR. The 

inclusion criteria were clinics with a high frequency of 

attendance by IIUM employees, used EMR and consented 

to the study. The exclusion criteria were those with low 

frequency of patients, used manual medical records and 

did not consent to the study. A total of four panel clinics 

were included, and data were collected from EMR. 

Patient selection 

The selection of patients was conducted using systematic 

sampling. Three patients per month were selected at the 

beginning, middle and end of each month from the 

PMCare database. The inclusion criteria were patients who 

had three or more medications. The search was conducted 

using a unique claim code that extracted the correct data 

regardless of the date recorded for database claim, 

patient’s visit or delayed submission of claims to PMCare. 

Data collection included information on claim code, date 

of prescription, name of the patient, patient identification, 

diagnosis, name of medications, number of medications 

per prescription, the number of drugs supplied, 

consultation fee, the total cost of medications and amount 

paid by the patient. Validated outcomes included 

diagnosis, medication name, number of medications per 

prescription, the quantity of medication, and the amount 

claimed to the health insurance. Majority of previous 

studies (83%) have reported similar data collection 

methods where cases were most frequently validated 

against diagnoses in medical records (Widdifield et al., 

2013). 

Accuracy measurement 

The accuracy was obtained by comparing the PMCare 

dataset with the original data from EMR by dividing the 

number of observations found to be the same by the total 

number of data elements of the PMCare database then 



Mohd Taufek et al. (2022) Journal of Pharmacy, 2(1), 31-35 

 

  

Page 33  

multiplying that total by 100. The agreement exceeding 

90% was considered as a good benchmark when data 

elements from the two sources were found not to be the 

same, noting the reasons (for example, unclear data 

definitions) as reported in previous studies (Iwamoto et al., 

2015; Miller et al., 2009). The discrepancy was determined 

when the inaccuracies of observations recorded in the 

EMR and PMCare database were identified. The 

discrepancies were calculated by dividing the number of 

observations found to be different with the total 

observations from selected PMCare data then multiplying 

by 100. 

 

Results 

 
A total of 118 patients (age: mean ±SD, 33.2 ± 12.7) were 

selected based on the availability of data from EMR (Table 

1). Although there were three patients selected per month 

from 2016-to 2019, many patients had to be excluded due 

to changes in the recording system used at the clinics, crash 

of the software and incomplete record of EMR. We found 

that several observations from EMR recorded extra 

medications (415 observations) compared to the PMCare 

database, which was 386 observations. Electronic medical 

records contained original data entered from patients’ 

prescriptions. The number of observations recorded in the 

EMR was higher than the PMCare database because the 

EMR captured the total number of medications prescribed 

to each patient. 

In contrast, only a few medications were recorded for 

insurance claims to PMCare for the coverage limit of 

RM45. Therefore, after removing extra observations from 

EMR for accuracy measurement, the total number of 

observations for both EMR and PMcare database was 386. 

As presented in Table 2, the number of patients included 

in the present pilot study  was sufficient as reported by a 

similar study that measured accuracy and validity of a 

hospital database (Cook et al., 2002). 

The advantage of using EMR was that additional 

information was available and comprehensively recorded 

for each visit. We found that consultation fees were varied 

between clinics and between clients depending on the 

number of medications prescribed and adjustment to the 

insurance eligibility.  Data from EMR provided the amount 

of out-of-pocket money paid by the patients if the total 

charges for consultation and medications exceeded the 

insurance coverage of RM 45, including the detailed price 

for each medication with a specific quantity. On the 

contrary, the PMCare database did not have this 

information. Discrepancies were found due to differences 

in recording the name and quantity of medications, 

diagnosis, the number of medications and the missing 

information in either EMR or PMCare database for the 

variables investigated. The most common diagnosis 

reported in the PMCare database was acute upper 

respiratory tract infection (URTI) (63.6%), followed by 

acute gastroenteritis and colitis (13.5%). 

 

 
Table 1: Demographics of patients from four private medical 

clinics. 

 

Characteristics N = 118 Frequency 

(%) 

Gender: 

a. Male 

b. Female 

 

63 

55 

 

53.4 

46.6 

Age (years): 

Mean (SD) 

Median (range) 

 

33.2 (12.7) 

35 (4 – 61) 

 

Diagnosis: 

a. Acute Upper Respiratory 

Infection, unspecified 

b. Infectious Gastroenteritis and 

Colitis, Unspecified 

c. Unspecified Abdominal Pain 

d. Anemia, Unspecified 

e. Others (Urticaria, Low back 

pain, Acute Tonsillitis, Viral 

Infection Unspecified, Dermatitis 

and Eczema, Conjunctivitis, 

Gout, Hemorrhoids, Hordeolum 

(Externum) (Internum) of Eyelid, 

Gastritis, Myalgia, Mycosis) 

 

75 

 

16 

 

3 

 

3 

21 

 

63.6 

 

13.5 

 

2.54 

 

2.54 

17.8 

 

Discussion 

The accuracy of the PMCare database has never been 

validated to determine its credibility to be used for 

database research in Malaysia. The present study reported 

that the PMCare claims database had overall 91.5% 

accuracy and 8.5% discrepancies compared with EMR 

data as a gold standard for four different panel clinics for 

the segregated patient samples collected in 2016, 2017 and 

2019. This outcome indicated an ideal accuracy value that 

was set as above 90% to be considered valid (Miller et al., 

2009). Our finding was lower than that reported  by a study 

in Japan, that demonstrated  99% accuracy in estimating 

opioid consumption among cancer patients (Iwamoto et 

al., 2015).  

Another study validated claims algorithm and 

clinical operative reports with 97% accuracy (Miller, 

Saigal & Warren, 2009). However, a different study has 

indicated a lower range of good accuracy with 80% and 

above for a pilot study reporting database validity (Cook 

et al., 2002). The analysis for an individual clinic showed 

that accuracy could achieve above 98% with a systematic 

and appropriate recording of claims data according to 

EMR (Table 2). 

 

 

 



Mohd Taufek et al. (2022) Journal of Pharmacy, 2(1), 31-35 

 

  

Page 34  

Table 2: Discrepancies of data from EMR of respective clinics 

in comparison to PMCare Database 

Variables Clinic 

A 

Clinic 

B 

Clinic 

C 

Clinic 

D 

 Total 

No. of patients 30 15 50 23 118 

Year of 

observations 

2016 2016 2016-

2017 

2019  

No. of 

Observations 

from EMR 

94 59 147 86 386 

No. of 

Observations 

from PMCare 

claims database 

94 59 147 86 386 

No. of 

observations with 

discrepancies 

7 9 2 15 33  

Types of 

discrepancies 

Name 

and 

quantity 

of 

medicat

ion. 

Name 

and 

quanti

ty of 

medic

ation, 

diagn

osis. 

Name 

of 

medicat

ion. 

Name 

and 

quantity 

of 

medicat

ion, 

diagnos

is. 

 

Percentage of 

discrepancies (%) 

7.4 15.3 1.4 17.4 8.5 

Percentage of 

accuracy (%) 

92.6 84.7 98.6 82.6 91.5 

 
Investigation of data discrepancy showed that types 

of discrepancy tend to vary between clinics. This could be 

due to the nature of PMCare claims data, where data were 

created and submitted later from the date of the patient’s 

visit, whereas the EMR data showed the actual 

medications dispensed daily. Although the differences in 

characteristics exist, both data were prone to human error 

in entering the details into the online system, which 

resulted in discrepancies (Iwamoto et al., 2015). It is also 

important to note that claims databases could only cater for 

specific data covered by the insurance company for certain 

medications and diseases. A review of previous studies has 

also highlighted the critical measures of diagnostic 

accuracy (sensitivity, specificity, and positive and negative 

predictive values) and disease prevalence in reporting 

validity of administrative database algorithms (Widdifield 

et al., 2013). 

Our study identified the discrepancies by including 

the missing names of medication or different medications 

recorded when compared between EMR and PMCare 

database. The diagnosis error was included when the 

diagnosis submitted to the PMCare database was different 

or unspecified compared to EMR. The limitation of this 

study was that we did not have specific details regarding 

the options given for each data category in the claim 

database system, such as general option (e.g. unspecified 

abdominal pain) or specific option for accurate diagnosis 

(e.g. dyspepsia, gastritis). We did not consider that 

specific, or general diagnosis of similar nature or location 

was different unless the difference was obvious such as 

abdominal pain (EMR) versus URTI (PMCare). From the 

perspective of the claims database and EMR, we propose 

that details listed in the claims database should be the same 

as those recorded in EMR. It was expected that the number 

of observations in EMR would be greater than those in the 

claim database since the insurance company covers only 

certain medications. The consistency in recording the data 

will improve the accuracy and validity of the data from 

both sources.  

The major limitation of our pilot study was that the 

sample size was relatively small compared to previous 

studies (Iwamoto et al., 2015; Takeda et al., 2016) but was 

sufficient to report validity for a pilot study (Cook et al., 

2002). Our study was multi-centred, with a standardised 

data format from the PMCare database, which allowed the 

comparison of varying practices between settings. Future 

studies may identify interventions for corrective actions to 

ensure high accuracy of the claims database. The results 

suggest that PMCare database is valid and accurate as a 

valuable resource to be used for research purposes but with 

careful analysis and interpretation considering the data 

characteristics. The identified discrepancies were common 

for certain variables in database validity studies as 

previously reported in other studies (Cook et al., 2002; 

Gabriel et al., 2005). However, improved documentation 

practice is required at the specific settings to improve the 

accuracy of the claims database. 

Conclusion 

Data from the health insurance claims database was valid 

and applicable for research purposes with careful analysis 

and interpretation. The overall high accuracy of the health 

insurance claims database in this study indicated its 

validity. However, variation in accuracy level was found 

between different clinics attributable to varying practices 

in recording the data. Future studies are required to 

identify effective interventions to improve the accuracy of 

claims databases with gold standards. 

Acknowledgement 

The authors would like to thank the PMCare database 

health insurance company and private panel health clinics 

of the International Islamic University Malaysia for their 

assistance in completing this research. CSZ was supported 

by a research grant from IIUM Health Wellness Centre. 

  



Mohd Taufek et al. (2022) Journal of Pharmacy, 2(1), 31-35 

 

  

Page 35  

Data Availability Statement 

Request to access the datasets should be directed to 

International Islamic University Malaysia Research 

Ethical Committee. The de-identified data could be shared 

with interested researchers after obtaining approval from 

the above ethical committee. The restriction on public data 

deposition is due to the privacy and confidentiality of 

patients’ health data. 

Conflict of Interest 

The authors have no conflict of interest to declare. The 

funder had no role in study design, data collection and 

analysis, decision to publish or prepare the manuscript. 

References 

Chua, H.T., & Cheah J.C. H. (2012). Financing universal 

coverage in Malaysia: a case study. BMC Public 

Health, 12(Suppl 1), S7. 

https://doi.org/10.1186/1471-2458-12-S1-S7 

Cook, S.F., Visscher, W.A., Hobbs, C.L., & Williams R.L. 

(2002). The Project IMPACT Clinical 

Implementation Committee Project IMPACT: 

Results from a Pilot Validity Study of a New 

Observational Database, Critical Care Medicine, 

30 (12), 2765-2770. doi: 10.1097/00003246-

200212000-00024 

Du, X.L., Key, C.R., Dickie, L., Darling, R., Geraci, J.M., 

& Zhang, D. (2006). External validation of 

Medicare claims for breast cancer chemotherapy 

compared with medical chart reviews. Med Care., 

44(2), 124–131. doi: 

10.1097/01.mlr.0000196978.34283.a6 

Gabriel, S.E., Amadio, P., & Ilstrup, D. (2005). The 

Feasibility and Validity Studies Comparing 

Orthopedists and Non-Orthopedists caring for 

musculoskeletal Injuries: Results of a Pilot study. 

Health Sciences Research, 10(3), 163-168. 

https://doi.org/10.1002/art.1790100303 

Iwamoto, M., Higashi, T., Miura, H., Kawaguchi, T., 

Tanaka, S., Yamashita, I., Yoshimoto, T., 

Yoshida S., & Matoba, M. (2015). Accuracy of 

using diagnosis procedure combination 

administrative claims data for estimating the 

amount of opioid consumption among cancer 

patients in Japan. Japanese Journal of Clinical 

Oncology, 45(11), 1036–1041. 

https://doi.org/10.1093/jjco/hyv130 

Koram, N., Delgado, M., Stark, J.H., Setoguchi, S., & de 

Luise, C. (2019). Validation studies of claims 

data in the Asia-Pacific region: A comprehensive 

review. Pharmacoepidemiology and Drug Safety, 

28(2), 156–170. https://doi.org/10.1002/pds.4616 

Miller, D.C., Saigal, C.S., Warren, J.L., Leventhal, M., 

Deapen, D., Banerjee, M., Lai, J., Hanley, J., & 

Litwin, M.S. (2009). External validation of a 

claims-based algorithm for classifying kidney-

cancer surgeries. BMC Health Services Research, 

9 (92). https://doi.org/10.1186/1472-6963-9-92 

Thomas, S., Beh, L., & Nordin, R.B. (2011). Health care 

delivery in Malaysia: changes, challenges and 

champions. Journal of Public Health in Africa, 

2(2), e23. https://doi.org/10.4081/jphia.2011.e23 

Tomlin, A.M., Reith, D.M., Woods, D.J., Lloyd, H.S., 

Smith, A., Fountain, J.S. & Tilyard, M.W. 

(2017). Pharmacoepidemiology database system 

for monitoring risk due to the use of medicines by 

New Zealand primary care patients. Drug Safety, 

40(12), 1259‐1277. 

https://doi.org/10.1007/s40264-017-0579-1 

Takeda, T., Mihara N., Murata T., Shimai Y., Okada K., 

Manabe, S., & Matsumura, Y. (2016). Estimating 

the ratio of patients with a certain disease 

between hospitals for the allocation of patients to 

clinical trials using health insurance claims data 

in Japan. Studies in Health Technology and 

Informatics, 228, 537‐541. Doi: 10.3233/978-1-

61499-678-1-537 

Widdifield, J., Labrecque, J., Lix, L., Paterson, J.M., 

Bernatsky, S., Tu, K., Ivers, N., Bombardier, C. 

(2013). Systematic review and critical appraisal 

of validation studies to identify rheumatic 

diseases in health administrative databases. 

Arthritis Care & Research, 65(9), 1490-1503. 

https://doi.org/10.1002/acr.21993 

Yamana, H., Moriwaki, M., Horiguchi, H., Kodan, M., 

Fushimi, K., & Yasunaga, H. (2017). Validity of 

diagnoses, procedures, and laboratory data in 

Japanese administrative data. Journal of 

Epidemiology, 10, 476–482. 

https://doi.org/10.1016/j.je.2016.09.009