1. Introduction
Healthcare overtime has become an important domain in the field of information technology due to the conjunction of technology and its use in different applications. Use of software technology seem becoming a need for management and processing of healthcare data [1, 2] and clinical documents [3]. Quality of healthcare is directly proportional to the quality of data stored in healthcare related software like EHRs, Billing Management Systems, Patient Health Records, Integrated Healthcare Systems [4-6].
Thousands of regular and periodic edits has made it slightly unmanageable for software vendors to cope with the changing requirements of health informatics domain [7]. Just use of software technology is not sufficient for improving quality of data and health informatics, underlying tools and techniques greatly impact the ability of software for proper management and processing of data [8]. This manuscript is for experts working in healthcare informatics domain. The manuscript discusses the underlying technologies of designing and developing healthcare related software. Two of the said technologies are compared in this study for the measurement of efficiency in implementation of medical coding/billing knowledge.
For healthcare IT professionals working in the domains of medical billing and computer programming, implementing medical edits sometimes becomes very complex. Medical edits include Mutually Exclusive Edits (MEE), Medically Unlikely Edits (MUE), Correct Coding Initiative (CCI) edits, Add-on Code Edits etc. The complexity of the task can be reduced by issuing specific instructions, related to coding and billing, given by individual providers. This research paper reports an important finding by comparing two underlying technologies of healthcare IT system development i.e. conventional programming technology and rule-based technology.
A brief overview of health informatics, issues and challenges is presented in the next section. A generic description of two technologies; rule-based expert systems’ technology and conventional software development technology, is then presented. In Section 2, Objectives of study have been presented. Section 3 is about the methodology of research. The results obtained during the research along with the statistical analysis are presented in Section 4 which is followed by the synthesis of paper in conclusion section presented at the end.
1.1. Health Informatics
This section presents snapshot of the needs and challenges of health informatics domain. Besides, user acceptance/adoptability, security and privacy of data, other major issues are data quality [6] and change flexibility which is the focus of this research study. Many organizations are working on these lines to overcome these issues. The organizations will be able to develop a more comprehensive health care system after addressing these issues and for which a cohesive Healthcare Information System is proposed [9]. For information and data sharing among programs, systems and institutions, all of the healthcare systems comprise of standardized set of procedures. Benefits of these healthcare systems include prompt information availability to providers, assurance of continuous care and inter-disciplinary communication [10].
Although multiple studies hypothesize the process of developing healthcare information systems yet only fewer offer the recommendations about it [11]. In [12], strategy method with a key element of ‘flexible standard’ is proposed for developing healthcare infrastructure. These flexible standards have two dimensions; ‘change flexibility’ and ‘use flexibility’. ‘Change flexibility’ refers to the changeability through modularization, and ‘Use flexibility’ refers to the magnitude with which multiple activities and tasks can be executed by a system. Moreover, some other issues related to medical billing processing are: 1. massive amount of heterogeneous data is to be processed, and 2. semantic aspects of data are to be considered, which implies that the information should be processed in such a way that it becomes readable and understandable by man and computer. This led us to use declarative knowledge representation (i.e. production rules) for implementation of healthcare edits.
Although, hundreds of EHRs, PHRs, and other healthcare information systems have been developed yet these systems cannot go beyond the capabilities of underlying conventional software development tools and techniques. Conventional programming technique cannot provide flexibility of change and use. This causes deterioration in quality of data with an exponential speed with the passage of time. The deterioration in the quality of data can ultimately effect patient safety and care [13].
Two techniques; conventional programming and rule-based programming have been scientifically compared in this research study. The following section presents a generic comparison of these two techniques, which will be followed by the hypothesis of the study.
1.2. Rule-Based Expert Systems a classical AI Technology
This section introduces the underlying technology of developing rule-based expert system, which is a classical technique of Artificial Intelligence, although initiated in 1970s in healthcare domain but nowadays frequently available in healthcare domain [14]. Rule-based programming is used to develop rule-based expert systems that can achieve change flexibility, hence can maintain quality of data with the passage of time [19], which ultimately contributes to better patient care.
Systems developed using rule-based programming comprises of three basic components namely knowledge base (a collection of ‘rules’), working memory and inference engine [15, 16]. Inference engines are those specified applications that are capable of reading a particular set of rules from the knowledge base, on input conditions pick the suitable rule and execute the conforming actions[14]. Rules are made up of if-then structures having diverse properties with respect to representation and computation. In rule-based programming, the term ‘rule’ (also known as ‘production rule’) refers to ‘if-then’ structure whereas ‘if’ represents condition and ‘then’ is followed by the action part.
IF
The production rule is written in the form of conventional if-then statement that is used as conditional control statement in the programming languages. The production rule, however, is based on the dataset and is applied onto the attributes of a database, that makes it different from the if-then used in programming languages [14, 17]. Coding in different programming languages requires specialized programming skills, time taken to execute a program, debugging and other incremental costs.
1.3. Conventional Software Development Technology
In conventional software systems, business rules (conditions and actions) are encoded in the form of compiled codes written in some programming language like Java, C++, C#, etc. With new innovations in healthcare new edits (conditions and actions) are required to be implemented on the monthly or quarterly basis, which in turn may induce new bugs in the software; and excessive time is wasted to eradicate those bugs. In conventional software ‘if’ statements are used (instead of production rules) as part of code and only persons with programming skills can change the code. Therefore, in case of conventional programming based software, professionals of healthcare domain are not able to incorporate updated knowledge in software by themselves. Some aspects of rule-based programming and conventional programming have been summarized in the Table 1.
Table 1. Comparison of Rule-Based programming and conventional programming
2. Objectives of the Study
The objectives of the case study are given as follows:
1. To compare rule-based technology and conventional software development technology for implementation of healthcare edits.
2. To study how the rule-based technology and conventional software development technology have been used for knowledge representation, utilization and rationalization in Health Informatics domain.
3. To familiarize healthcare professionals with underlying software technologies being used in management of healthcare data.
2.1. Hypothesis
The Null hypothesis is that, there is no difference of time in implementation of medical billing edits using rule based programming and using conventional programming. The hypothesis is constituted only for objective 3 since that is the only assessable and measureable objective listed in section 2.
The following hypothesis (equation 1 and 2) was developed for the study.
Ho (Null Hypothesis, µR- µC =0 or µR=µC (1)
H1 (Alternate Hypothesis, µR- µC ≠0 or µR≠µC (2)
Where µR denotes the average implementation time of system checks (healthcare edits) through rule-based programming and µC denotes the average implementation time of system checks (healthcare edits) through conventional programming. Implementation time depicts total time spent on each check from analysis to coding and making it operational. In the next section methodology of research has been described.
3. Proposed Method
In order to meet objective 2 and 3 of the study (as given in Section 2), we propose a method of medical claim processing that has been selected as application domain to test the above mentioned research hypothesis, mainly because it is a knowledge rich domain, where thousands of rules and regulations have been defined. Advantage in healthcare management domain is saving providers’ precious time from data management related activities.
3.1 Search Strategy and Study Selection
This research proposed a mechanism for implementing rule-based healthcare edits. Rule-based healthcare edits are claimed to improve the performance of managing healthcare edits with respect to time and human error. Two teams (each comprising of two members) with approximately same skills in software development were assigned task of implementing the rules and regulations related to medical claim processing. Both the members of the team were having expertise skills in software development. The participants are working in a medical billing company in software development section. One team was using conventional software development technology technique and the other team was using rule-based technology to develop the software. Time taken by both the teams for encoding of rules and regulations of medical claim processing was recorded as major part of this research study.
3.2 Data Sources
Medical edit is not a technical term. It is devised for this paper to encapsulate the concept of changes in medical claims that are asked by the user of medical service. This term is used in medical claim billing. In this study, a sample of 100 medical billing edits was selected; which were implemented simultaneously by the two teams; one using conventional software development technology and other team using rule-based technology. The data depicted time (in hours: minutes) consumed to implement particular edit through rule-based technology and conventional software development technology consecutively.
3.3 Inclusion Criteria
An edit is requested after submission of a medical claim. For an edit to be processed, it should qualify a predefined criterion. Prior to the implementation of rule-based technology, the conventional software development technology was being used for implementation of edits. During this research multiple phases of edit processing and inclusion were practically studied. The inclusion of edit depends upon the domain of edit, its alignment with the rules, time of request and amount of claim. In this study, critical evaluation of 100 edits processed by conventional systems and rule-based technology are considered.
3.4 Exclusion Criteria
Complexity of the edits is the major criteria for exclusion from implementation stage of this research study. Therefore, the edits like MEE (Mutually Exclusive Edits), NCCI (National Correct Codding Initiative) have not been included as it will require more complex programming for their implementation.
3.5 Analysis of Data
To analyze the above-mentioned hypothesis Statistical Package for Social Science (SPSS) with the latest version (V24.0) was applied. SPSS is the modern analytical approach and user-friendly software. It allows multiple kinds of analysis, the transformation of data and forms of output. The following statistical techniques were applied:
Paired t-Test: To compare two different programming techniques on the same checks Paired t-test was applied [18]. Paired T-Test is used to find the difference between two variables that are used on the same subject. The subject variables are projected on n different time intervals with all the other relative information as same. In this study, a total 100 system checks were selected and implementation time of each check by rule-based programming and conventional programming was compared.
Cronbach's Alpha: To measure the internal consistency of sample set as a group Cronbach’s alpha [19] was used. Cronbach’s alpha is a measure of scale reliability. To calculate the inter correlation among the two test items namely rule-based programming and conventional programming. The standardized formula of Cronbach’s Alpha is given in equation 3: (3)
\(\alpha=\frac{N \cdot c^{-}}{v^{-}+(N-1) \cdot c^{-}}\) (3)
Where N refers to the number of sample items; refers to the average covariance between sample items, and refers to the average variance of each sample item.
4. Result
Table 2 shows the time taken by the implementation teams in production rule-based programs and conventional if-then programs. The test is based on 100 edit implemented using both the teams respectively. In Table 2, the column ‘As Rule’ shows the implementation time of the edit in the form of Production-Rule, while the column ‘As IF-Then’ shows the implementation time as If-Then statement using conventional programming technology.
Table 2. Time to execute rule-based and conventional if-then methods
5. Analysis and Discussion
First part of this section presents empirical analysis of results, while in order to scientifically confirm the authenticity of results, statistical analysis has been given in the later part of this section. In Table 2, column ‘Sr’ is the order according to which the production rules were applied to validate the checks. The checks are dependent on each other and those applied in lateral stage are effected by the ones applied earlier. These cross-dependencies of production rules caused increase in implementation time. The complexity of debugging also increased with increased depth of inheritance tree and number of functions. On the other rule, the complexity of production rules remains constant and will not be affected by increasing these cross-dependencies to any extent.
The results empirically have shown that the time spent on programming and implementing the billing edits is almost thrice to the same implemented through production rules (91 Hours vs. 254.5 Hours). This means, that the use of rule based systems increases the performance of a healthcare professional who can now design and test billing edits three times faster than the conventional application programs written by the programmers.
Moreover, it is observed that the edits implemented earlier in sequence are affecting the edits implemented later. As discussed earlier, in addition to this, the time taken in implementing such modules with higher depth of inheritance tree and cyclomatic complexity, the coding becomes increasingly difficult so as the debugging. Production rule development, on the other hand absorbs the effect of increase in the size of edits and won’t show any remarkable increase in time taken to complete the execution. Fig. 1, given below depicts the implementation curves of both techniques graphically.
Fig. 1. Comparison of implementation time of healthcare edits
Rule-based technology has the linear implementation curve while the conventional technology is not linear in increase. This implies that with the increase in number of edits conventional technology may fail very soon. By making this comparison objective 1 of the study is also achieved.
In order to perform statistical analysis SPSS Variable sheet was developed showing two variables namely time (in minutes) to implement edit by rule-based technology and time (in minutes) to encode/implement the same edit by conventional software development technology. Then the data of paired variables was imported to SPSS data sheet. Data was analyzed using two-tailed t-tests, Cronbach’s alpha. The results and interpretation of the data analysis are as follows:
5.1 Paired t-Test
Two-tailed t-test at 95% level of significance was conducted to compare means of a sample of 100 checks. The sample of 100 was taken so that it could represent the whole population in an accurate manner.
The results in Table 3 depict that to implement a rule involving all step analysis, the definition of the rule, testing of the rule and moving the rule to production, on average 54.60 or approximately 55 minutes are required; whereas to update/write code, 152.75 or approximately 152.75 minutes are required.
Table 3. Development time stats of paired inputs
Hence implementation of checks by conventional programming takes 3 times more time than the implementation of healthcare edits by rule-based technology.
Ho (Null Hypothesis, µR- µC =0
Or µR=µC
Where µR represents average time or arithmetic mean to develop system check by rule-based programming and µC represents average time to implement system check by conventional programming. The Table 4 shows that the null hypothesis (stating that the equal average time is required to apply system checks by rule-based programming and conventional programming) is rejected at 95% level of significance.
Table 4. Paired Samples Test
5.2 Reliability
To measure the internal consistency or reliability of the sample items, Cronbach’s alpha was administered, depicted by Table 5. The reliability ranges from 0 to 1.
Table 5. Case Processing Summary
The value of Cronbach’s Alpha for 2 items is 0.462 which is a low level of internal consistency for this sample. As α ≤ 0.5 depicts that sample items are independent of one another so α=0.462 shows that time to implement system check by rule-based programming is independent of time to implement system check by conventional programming. High covariance between each pair leads to high variation among the total time consumed to execute system checks by rule-based programming and conventional programming.
The statistics specifies that rule-based programming is three times more time efficient than conventional programming. The data analysis showed that the hypothesis stating that on average equivalent time was used to implement system check by conventional programming and rule-based programming was rejected at 95% level of significance. Moreover, as Cronbach’s alpha value is less than 0.5 so both the samples are independent or in other words time taken by conventional programming is independent of time take by rule-based programming. High covariance in each pair of the paired sample led to the high total variation between both samples.
6. Conclusion
Two technologies have been scientifically compared for the implementation of healthcare edits. Empirical analysis – authenticated statistically – has proved that rule-based technology is approximately 3 times more efficient than conventional software development technology. The results of the study are valuable when comparing either, pre and post implementation of rule-based technology or comparing two systems utilizing rule-based technology and conventional software development technology. The reason for which rule based systems outperformed others is their flexibility in approach and simplicity in managing the production rules. The rules are extracted from the existing datasets and to add, remove or modify a rule is not a design time issue. Rules can be edited at runtime by editing the relevant data only without making any change to the source code. Similarly, rule based programming is suitable when a large number of edits based on knowledge are expected within a dataset.
References
- Tegegne, E. Workineh, P. Seppanen, and M. O. Ahmad, "Software development methodologies and practices in start-ups," IET Software, vol. 13, no. 6, pp. 497-509, 2019. https://doi.org/10.1049/iet-sen.2018.5270
- F. S. Ujager and A. Mahmood, "A context-aware accurate wellness determination (caawd) model for elderly people using lazy associative classification," Sensors, vol. 19, p. 1613, 2019. https://doi.org/10.3390/s19071613
- S. Hussain, M. Hussain, M. Afzal, J. Hussain, J. Bang, H. Seung, and S. Lee, "Semantic preservation of standardized healthcare documents in big data," International journal of medical informatics, vol. 129, pp. 133-145, 2019. https://doi.org/10.1016/j.ijmedinf.2019.05.024
- C. Dainton and C. H. Chu, "A review of electronic medical record keeping on mobile medical service trips in austere settings," International journal of medical informatics, vol. 98, pp. 33-40, 2017. https://doi.org/10.1016/j.ijmedinf.2016.11.008
- K. A. Kelley, H. E. Hoops, L. Palmer, N. A. Cohen, and K. J. Brasel, "Implementation of a medical coding curriculum for surgery residents," The American Journal of Surgery, vol. 217, pp. 834-838, 2019. https://doi.org/10.1016/j.amjsurg.2019.02.027
- M.-L. Sher, P. C. Talley, T.-J. Cheng, and K.-M. Kuo, "How can hospitals better protect the privacy of electronic medical records? Perspectives from staff members of health information management departments," Health Information Management Journal, vol. 46, pp. 87-95, 2017. https://doi.org/10.1177/1833358316671264
- D. Brahams, "A Claim Delayed May Be a Claim Denied," Medico-Legal Journal, vol. 75, no. 2, pp. 37-39, 2007. https://doi.org/10.1258/rsmmlj.75.2.37
- V. Mantzana, M. Themistocleous, Z. Irani, and V. Morabito, "Identifying healthcare actors involved in the adoption of information systems," European Journal of Information Systems, vol. 16, pp. 91-102, 2007. https://doi.org/10.1057/palgrave.ejis.300066
- T. Maenpaa, T. Suominen, P. Asikainen, M. Maass, and I. Rostila, "The outcomes of regional healthcare information systems in health care: a review of the research literature," International journal of medical informatics, vol. 78, pp. 757-771, 2009. https://doi.org/10.1016/j.ijmedinf.2009.07.001
- P. M. Ginter, W. J. Duncan, and L. E. Swayne, The strategic management of health care organizations, John Wiley & Sons, 2018.
- O. Hanseth and K. Lyytinen, "Design theory for dynamic complexity in information infrastructures: the case of building internet," Enacting research methods in information systems, Springer, pp. 104-142, 2016.
- I. Bowns, G. Rotherham, and S. Paisley, "Factors associated with success in the implementation of information management and technology in the NHS," Health Informatics Journal, vol. 5, pp. 136-145, 1999. https://doi.org/10.1177/146045829900500305
- Y.K. Alotaibi, SBFM. ABFM, and F. Federico, "The impact of health information technology on patient safety," Saudi Med Journal, vol. 38, pp. 1173-1180, 2017. https://doi.org/10.15537/smj.2017.12.20631
- U. Abdullah, Data mining driven learning apprentice system for medical billing compliance, Foundation University, Islamabad, FUIEMS. Islamabad, Pakistan: Higher, 2012.
- A. K. Akanbi, M. Masinde, "Towards the Development of a Rule-Based Drought Early Warning Expert Systems Using Indigenous Knowledge," in Proc. of 2018 International Conference on Advances in Big Data, Computing and Data Communication Systems, pp. 1-8, 2018.
- S. Chaudhari, V. Sardar, D.S. Rahul, M. Chandan, M.S.Shivakale, and K.R.Harini, "Performance Analysis of CNN AlexNet and VGGNet Models for Drought Prediction using Satellite Images," in Proc. of Innovation in Technology (ASIANCON) 2021 Asian Conference, pp. 1-6, 2021.
- U. Abdullah, A. Ligeza, and K. Zafar, "Performance evaluation of rule-based expert systems: An example from medical billing domain," Expert Systems, vol. 34, p. e12218, 2017. https://doi.org/10.1111/exsy.12218
- N. Kock, "One-tailed or two-tailed P values in PLS-SEM?," International Journal of e-Collaboration (IJeC), vol. 11, pp. 1-7, 2015.
- L. J. Cronbach, "Coefficient alpha and the internal structure of tests," psychometrika, vol. 16, pp. 297-334, 1951. https://doi.org/10.1007/BF02310555
- M. Shaheen, U. Abdullah, "CARM: Context Based Association Rule Mining for Conventional Data," Computers, Materials & Continua, vol. 68, 2021.