Evaluation of the International Classification of Health Interventions (ICHI) in the coding of common surgical procedures

Abstract

Objective

To evaluate the International Classification of Health Interventions (ICHI) in the clinical and statistical use cases.

Materials and Methods

We identified 300 most-performed surgical procedures as represented by their display names in an electronic health record. For comparison with existing coding systems, we coded the procedures in ICHI, SNOMED CT, International Classification of Diseases (ICD)-10-PCS, and CCI (Canadian Classification of Health Interventions), using postcoordination (modification of existing codes by adding other codes), when applicable. Failure analysis was done for cases where full representation was not achieved. The ICHI encoding was further evaluated for adequacy to support statistical reporting by the Organisation for Economic Co-operation and Development (OECD) and European Union (EU) categories of surgical procedures.

Results

After deduplication, 229 distinct procedures remained. Without postcoordination, ICHI achieved full representation in 52.8%. A further 19.2% could be fully represented with postcoordination. SNOMED CT was the best performing overall, with 94.3% full representation without postcoordination, and 99.6% with postcoordination. Failure analysis showed that “method” and “target” constituted most of the missing information for ICHI encoding. For all OECD/EU surgical categories, ICHI coding was adequate to support statistical reporting. One OECD/EU category (“Hip replacement, secondary”) required postcoordination for correct assignment.

Conclusion

In the clinical use case of capturing information in the electronic health record, ICHI was outperformed by the clinically oriented procedure coding systems (SNOMED CT and CCI), but was comparable to ICD-10-PCS. Postcoordination could be an effective and efficient means of improving coverage. ICHI is generally adequate for the collection of international statistics.

INTRODUCTION

The International Classification of Diseases (ICD) has been in use for the collection of global health trends and statistics of diseases for over a century.^1,2 However, there is no equivalent international classification standard for medical and surgical procedures.³ In 1976, the year after the official release of ICD-9, the World Health Organization (WHO) published the International Classification of Procedures in Medicine (ICPM).⁴ ICPM covered procedures for medical diagnosis, prevention, therapy, radiology, drugs, and surgical and laboratory procedures.⁵ Work on ICPM stopped in 1989 when the WHO decided not to revise ICPM in conjunction with ICD-10, because “the process of consultation that had to be followed before finalization and publication was inappropriate in such a wide and rapidly advancing field.”⁶ Almost 2 decades later, in 2007, work on an international procedure classification recommenced. The WHO decided to broaden the scope to include all health interventions and called this new classification International Classification of Health Interventions (ICHI).^7–11 The WHO defines health intervention as “an act performed for, with or on behalf of a person or a population whose purpose is to assess, improve, maintain, promote or modify health, functioning or health conditions.” ICHI covers interventions carried out by a broad range of providers across the full scope of health systems and includes diagnostic, medical, surgical, mental health, primary care, allied health, functioning support, rehabilitation, traditional medicine, and public health interventions. After 3 alpha releases, the first beta version of ICHI was released in 2017. ICD, International Classification of Functioning, Disability and Health (ICF), and ICHI constitute the 3 reference classifications within the WHO Family of International Classifications (WHO-FIC), which is a set of integrated classifications that provide a common language for health information across the world. Currently, ICHI is in its beta-3 release and is available for browsing online through the unified browser platform for the 3 WHO-FIC reference classifications.¹² The medical and surgical interventions in the current version of ICHI are considered to have a stable structure and are ready for implementation.

ICHI provides governments, service providers, managers, and researchers with a common tool for reporting and analyzing health interventions for statistical, quality, and reimbursement purposes.¹³ Internationally, ICHI enables comparison of data between countries, which will support important WHO initiatives such as Universal Health Coverage—the provision of essential health interventions to all people, such as antenatal care, measles vaccination, and hypertension treatment. Nationally, ICHI can be used in several ways. For countries that do not have a procedure classification system, they can use ICHI directly, or develop their national system based on ICHI. For countries that already have a procedure classification system, they can extend their system by incorporating ICHI’s broader range of interventions. Moreover, they can map their own system to ICHI to enable comparison with other countries.

The scope of ICHI is quite broad, ranging from surgical procedures to functional support, rehabilitation, and public health interventions. A literature search revealed studies that evaluated ICHI in the domains of audiology,¹⁴ nursing,¹⁵ disability,¹⁶ and public health interventions.¹⁷ These domains probably attract special attention because there are not covered by any international classifications before ICHI and there are very few competing coding systems. However, in the domain of surgical procedures, ICHI overlaps with existing national and international coding systems. Surgical procedures are an important component of healthcare. It has been estimated that approximately a third of the global disease burden requires surgical and/or anesthetic care.¹⁸ There are relatively few studies about ICHI involving surgical procedures. There are some studies that compared ICHI to ICD-9-CM procedure codes, but ICD-9-CM is no longer in active use.^10,11 Another study compared ICHI with SNOMED CT, covering only a small subset of surgical procedures.¹⁹

In this study, we focus on surgical procedures and compare ICHI to 3 national and international procedure coding systems in their ability to encode surgical procedures in the electronic health record. A comparison between ICHI and other systems will help to define the role that ICHI can play among surgical procedure coding systems. In addition, we assess the ability of ICHI to support the collection of international statistics on surgical procedures. Here, we report our results and discuss how ICHI can be enhanced to improve coverage. We believe ours is the first systematic comparison of ICHI and other current coding systems for the encoding of surgical procedures.

MATERIALS AND METHODS

Procedure coding systems

The 4 procedure coding systems evaluated in this study are ICHI, SNOMED CT, ICD-10-PCS (ICD-10 Procedure Coding System), and CCI (Canadian Classification of Health Interventions). All 4 systems have main codes (also called “precoordinated codes” or “stem codes”) that can be used on their own. ICHI, SNOMED CT, and CCI allow users to optionally add other codes to the main codes to modify or refine their meaning. This is often referred to as “postcoordination.” Note that the use of multiple main codes together is not postcoordination, because there is no change in the meaning of the main codes. The following is a brief description of each system (Table  1).

Table 1.

Characteristics of the 4 procedure coding systems

	ICHI	SNOMED CT	ICD-10-PCS	CCI
Semantic structure	3 axes—target, action, means	Concepts defined by attributes and values based on the SNOMED CT concept model	7 axes—section, body system, root operation, body part, approach, device, qualifiera	6 axes—section, group, intervention, qualifier 1 (approach, technique, reason, method, route, nature of administration), qualifier 2 (device, agent, method, tool used), and qualifier 3 (tissue used, type, group, strain)
Example	KCF.JK.AB Laparoscopic cholecystectomy KCF = gall bladder (target); JK = excision, total (action); AB = percutaneous endoscopic(means)	45595009 Laparoscopic cholecystectomy defined by: method = excision-action; procedure site-direct = gallbladder structure using access device = laparoscope, device	0FT44ZZ Resection of gallbladder, percutaneous endoscopic approach 0 = medical and surgical (section); F = hepatobiliary system and pancreas (body system); T = resection (root operation); 4 = gallbladder (body part); 4 = percutaneous endoscopic (approach); Z = no device; Z = no qualifier	1.OD.89. DA Excision total, gallbladder endoscopic [laparoscopic] approach without extraction (of calculi) cholecystectomy alone 1 = Physical/Physiological Therapeutic Interventions; OD = gallbladder; 89 = excision total; DA=endoscopic (laparoscopic) approach
Number of codes: a. total; b. surgical proceduresb (version)	6862 stem codes; 4931 stem codes (2020 Beta-3 release)	354 384 codes; 20 137 codes (July 2020 release)	77 559 codes; 67 257 codes (2020 release)	17 425 codesc 16 010 codes (2018 release)
Postcoordination	Yes	Yes	No	Yes (optional attributes)
Use in	WHO member countries	SNOMED International member countries and affiliates	United States, Belgium, Spain, Portugal, and others	Canada

CCI: Canadian Classification of Health Interventions; ICD: International Classification of Diseases; ICHI: International Classification of Health Interventions; WHO: World Health Organization.

^aThe ICD-10-PCS axes can be different depending on the section, this list of axes applies to the largest section of Medical and Surgical Procedures.

^bThe number of surgical procedure codes is an approximation and can include nonsurgical codes. It is estimated as follows:  • ICHI—all codes from Chapter 1: Interventions on Body Systems and Functions  • SNOMED CT: codes under the Surgical procedure subhierarchy  • ICD-10-PCS: all codes under Medical and Surgical Procedures  • CCI: codes in Section 1 Physical/Physiological Therapeutic Interventions, Section 2 Diagnostic Interventions, and Section 5 Obstetrical and Fetal Interventions

^cCounts reflect main codes only, not considering mandatory or optional attributes.

International Classification of Health Interventions

ICHI is based on 3 axes, each forming a part of the 7-character code. The axes are:

• target—the entity on which the action is carried out

• action—the deed done by an actor to the target

• means—the processes and methods by which the action is carried out.

ICHI allows postcoordination, which means that the main ICHI codes (called “stem codes” when they are used for postcoordination) can be refined by combination with extension codes to add additional detail.

SNOMED CT

SNOMED CT is the emerging international clinical terminology standard.^20,21 SNOMED International has 41 member countries and has issued affiliate licenses to more than 5000 individuals and organizations.²² SNOMED CT covers most domain areas relevant to clinical medicine (eg, diseases, procedures, drugs) and the surgical procedure subhierarchy is the focus of this study. SNOMED CT concepts are defined logically by attributes and values following a concept model.²³ Existing concepts can be refined with postcoordination by adding attributes and values in accordance with the concept model.

ICD-10-PCS

ICD-10-PCS was created in the United States as a replacement for the procedure codes in ICD-9-CM (retired in 2015).²⁴ The main use of ICD-10-PCS in the United States is for classifying procedures performed in hospital inpatient healthcare settings.²⁵ Outside the United States, ICD-10-PCS is also used in other countries, such as Belgium, Spain, and Portugal. ICD-10-PCS has a 7-character code, with each character representing an axis of classification. While multiple ICD-10-PCS codes can be used to describe a procedure, there is no option to modify the meaning of an existing code by postcoordination.

Canadian Classification of Health Interventions

CCI was introduced in Canada in April 2001 with a staggered implementation and was widely adopted by 2004–2005.²⁶ CCI was developed and is maintained by the Canadian Institute for Health Information (CIHI). CCI is based on the International Standard (ISO 1828) for classifying surgical interventions that allow for international mapping and comparability.²⁷ The initial development of ICHI was also based on the same ISO standard. CCI is used across the continuum of healthcare settings in Canada. It has a 3-year update cycle to ensure it remains relevant. CCI is based on 6 axes of which the last 3 are qualifiers. Additional data elements extraneous to the CCI code (called “attributes”) provide extra detail about status, location, or extent of the intervention where applicable. Depending on the specific code, attributes can be mandatory (always required) or optional (user will decide whether to add or not). In our study, we consider optional attribute as a kind of postcoordination.

The characteristics of the 4 systems are summarized in Table  1. Examples of postcoordination can be found in Table  2 under “Results.”

Table 2.

Encoding common surgical procedures in 4 procedure coding systems

	ICHI	SNOMED CT	ICD-10-PCS	CCI
Full representation without postcoordination	121 (52.8%)	216 (94.3%)	167 (72.9%)	204 (89.1%)
—Single code	—98 (42.8%)	—201 (87.8%)	— 62 (27.1%)	—196 (85.6%)
—Multiple codes	—23 (10%)	—15 (6.6%)	—105 (45.9%)	—8 (3.5%)
Example: Procedure name → Code(s)	Cholecystectomy → KCF.JK.AA Cholecystectomy	Cholecystectomy → 38102005 Cholecystectomy	Cholecystectomy→ 0FT40ZZ Resection of Gallbladder, Open Approach	Cholecystectomy→ 1.OD.89. LA Excision total, gallbladder open approach
Full representation with postcoordination	44 (19.2%)	12 (5.2%)	NA	7 (3.1%)
Example: Procedure name → Code(s) + postcoordinated expression	Amputation of toe at metatarsophalangeal joint → PNE.JN.AA Amputation of toe, not elsewhere classified and XA8XU1 Metatarsophalangeal joint	Laparoscopic destruction of intra-abdominal endometriomas → 63831004 Destruction of intra-abdominal endometriomas + Using access device = Laparoscope, device		Excision Bartholin’s cyst → 1.RW.87.LA Excision partial, vulva using open approach and no tissue + optional location attribute GV—Greater vestibular glands [Bartholin's]
Partial representation	64 (27.9%)	1 (0.4%)	62 (27.1%)	18 (7.9%)
Example: Procedure name → Code(s)	Capsulotomy, YAG laser → BBF.FA.AA Capsulotomy	Arthroplasty and implant elbow with reconstruction fascia lata ligament → 239471004 Arthroplasty of the elbow (procedure)	Cryosurgery Cervix → 0U5C7ZZ Destruction of Cervix, Via Natural or Artificial Opening	Hernia repair, inguinal→ 1.SY.80.LA Repair, muscles of chest and abdomen open approach without tissue (eg, suturing or stapling)
Total	229 (100%)	229 (100%)	229 (100%)	229 (100%)

CCI: Canadian Classification of Health Interventions; ICD: International Classification of Diseases; ICHI: International Classification of Health Interventions.

Encoding of common surgical procedures in 4 procedure coding systems

We utilized the statistics of surgical procedures provided by a large US healthcare provider for a previous project.²⁸ We identified the top 300 most-performed procedures and encoded each procedure in the 4 coding systems based on their display names in the electronic health record. We ignored some extraneous information in the display names, which was considered out-of-scope for procedure coding systems. Examples of extraneous information included patient age (eg, Tonsillectomy and adenoidectomy up to age 12), indication (eg, Gastric bypass for morbid obesity), and timing (eg, rotator cuff repair, open, acute). If there was no explicit mention of surgical approach (eg, open, laparoscopic, transluminal), we assumed an open approach. We assumed complete (as opposed to partial) removal of a body part when either was possible. After removing duplicates, 229 unique procedures remained. Encoding was done independently by 2 terminologists who were experienced with the particular coding system. All discrepancies were discussed until consensus was reached.

For each procedure, we looked for the best-matching code, which was either equivalent or broader in meaning than the display name of the procedure. More than one code could be used if the combination of codes matches the meaning of the procedure. If broader codes were used, we would attempt postcoordination in ICHI, SNOMED CT, and CCI to refine the code and improve the matching. We defined 3 levels of matching: (1) full representation without postcoordination, (2) full representation with postcoordination, and (3) partial representation.

Specifically, the following describes the coding process in each system.

International Classification of Health Interventions

We used the online browser provided by the University of Udine²⁹ (the unified WHO browser was not available then) and followed the corresponding coding guidelines and instructions. We considered all relevant information (target, action, means, ICHI descriptor, definition, index terms, and include notes) when deciding on the ICHI code. To determine the degree of representation, we considered only the information in the ICHI code descriptor and the 3 axes. For example, the procedure “Capsulotomy, YAG laser” was coded as BBF.FA.AA Capsulotomy (target = lens, action = incision, means = open approach). Even though BBF.FA.AA had an inclusion “Yttrium-aluminum-garnet [YAG] laser,” we still considered this partial representation because “laser” was not mentioned in the descriptor or the axes. The use of multiple stem codes was allowed. For example, the procedure “rhinoplasty with septoplasty” was coded as a combination of JAA.ML.AA Reconstruction of nose and JAB.ML.AC Septoplasty. ICHI has its own syntax for combination of codes. When multiple procedures are performed together, the stem codes should be separated by “/” (ie, JAA.ML.AA/JAB.ML.AC in the example above). For procedures that were not fully represented with main ICHI codes, we would attempt postcoordination to achieve full representation. In postcoordination, the stem code is separated from the extension code by “&.”

SNOMED CT

We used the online browser provided by SNOMED International.³⁰ We judged the degree of representation based on the fully specified name, hierarchical position, and defining attributes of the SNOMED CT concept. We used multiple precoordinated concepts if necessary. For example, “Hemorrhoidectomy, internal and external, simple” was coded as 22432007 Internal hemorrhoidectomy and 61498008 Complete external hemorrhoidectomy. We would attempt postcoordination when precoordinated concepts could not achieve full representation.

ICD-10-PCS

We used the MAGPIE (Map-Assisted Generation of Procedure and Intervention Encoding) browser that we created and is publicly available from the National Library of Medicine website.³¹ MAGPIE helps users find ICD-10-PCS codes through the combination of lexical and code-based matching, leveraging the rich ICD-9-CM procedure codes index and resources such as the UMLS and other existing maps.^32,33 We allowed multiple ICD-10-PCS codes if necessary. For example, “Cesarean section” was coded as 10D00Z2, 10D00Z1, or 10D00Z0, corresponding to Open extraction of products of conception via: extraperitoneal, low, or high approach, respectively. Postcoordination is not an option in ICD-10-PCS.

Canadian Classification of Health Interventions

We used the advance query search option in the CCI 2018 Folio viewer for encoding. In CCI, additional detail is available as attributes (eg, status, location, extent, and mode of delivery), which can be mandatory or optional to assign. Mandatory attributes were considered part of the existing (precoordinated) codes. We permitted the assignment of multiple codes when necessary. If precoordinated codes could not achieve full representation, addition of optional attributes would be attempted if they were available. The use of optional attributes was considered postcoordination.

Failure analysis

All cases that achieved partial representation were reviewed to determine what kind of information was missing. We categorized the type of missing information according to the 3 ICHI axes—target, action, or means. We further subdivided “means” into 3 subcategories: method (eg, laser, cryosurgery), approach (eg, retropubic), and device (eg, elbow implant).

Use of ICHI in international statistics

The Organisation for Economic Co-operation and Development (OECD) and European Union (EU) currently report data from member countries on hospital interventions, including a common list of 22 surgical procedure categories (eg, cataract surgery, tonsillectomy, and laparoscopic appendectomy).^34–36 We assessed the adequacy of ICHI codes to support OECD and EU statistical reporting as follows. Among the 229 surgical procedures that were coded in ICHI, we identified those that were encompassed in the OECD/EU list of surgical interventions. We then assessed whether the ICHI encoding of these procedures contained sufficient information for accurate and unambiguous assignment to their corresponding OECD/EU categories.

RESULTS

Encoding of common surgical procedures in 4 procedure coding systems

Coding in each system was done by 2 terminologists independently. Before coding, the terminologists worked together on some examples to establish coding guidelines. Actual coding started when they were satisfied with the guidelines. The intercoder agreement for the main codes, before discrepancies were discussed, was ICHI 91.7%, SNOMED CT 85.6%, ICD-10-PCS 72.9%, and CCI 97.8%. The relatively low agreement in ICD-10-PCS was related to the high incidence of the need for multiple codes which increased the chance for discordance (see below).

The results of encoding are summarized in Table  2. Overall, SNOMED CT had the best performance. Without postcoordination, SNOMED CT was able to achieve full representation in 216 (94.3%) procedures. With postcoordination, full representation was achieved in all but one procedure. Without postcoordination, ICHI achieved full representation in 52.8%, lower than ICD-10-PCS (72.9%) and CCI (89.1%). With postcoordination, ICHI was able to match ICD-10-PCS, achieving full representation in 72%.

Another interesting finding was that, among the procedures that could be fully represented without postcoordination, the proportions of procedures that required multiple codes to represent were 10%, 6.6%, 45.9%, and 3.5% in ICHI, SNOMED CT, ICD-10-PCS, and CCI, respectively. As an example, “breast augmentation” can be fully represented by a single code in SNOMED CT (22890008 Augmentation mammoplasty), CCI (1.YM.79.LA-PM Repair by increasing size, breast open approach without tissue with implantation of prosthesis), and ICHI (LCA.LK.AA Augmentation mammoplasty, not elsewhere classified). However, in ICD-10-PCS, there are 12 possible codes, depending on laterality (left, right, bilateral) and the device used (autologous tissue substitute, synthetic prosthesis, etc.). This shows that the ICD-10-PCS codes are generally finer-grained (more detailed) than the level at which surgical procedures are recorded in the electronic health record. On the other hand, the granularity of SNOMED CT and CCI aligns better with clinical documentation.

Failure analysis

The reasons for the failure of full representation are summarized in Table  3. The percentages in each type of missing information are based on the number of procedures. The individual percentages can add up to over 100% since one procedure can have more than one type of missing information. For both ICHI and ICD-10-PCS, “method” was the most common missing information. For CCI, missing “target” occurred in over 60% of cases. The only case of partial representation in SNOMED CT was due to a missing method.

Table 3.

Types of missing information in procedures that are partially represented (the numbers represent the number of procedures with a particular type of missing information, the percentages for each type of missing information are calculated using the number of procedures with missing information for that classification as the denominator, the percentages can add up to over 100% because one procedure can have multiple types of missing information)

Type of missing information	ICHI	SNOMED CT	ICD-10-PCS	CCI
Target	21 (32.8%)		24 (38.7%)	11 (61.1%)
Example: Procedure name Code found Missing information	Transposition of ulnar nerve at elbow ACA.LJ.AA Transposition of peripheral nerve Ulnar nerve		Repair of umbilical hernia 0WQF0ZZ Repair Abdominal Wall, Open Approach Umbilical hernia	High ligation and stripping of long saphenous vein 1.KR.87.WM Excision partial, veins of leg NEC Long saphenous vein
Action			2 (3.2%)
Example: Procedure name Code found c. Missing information			Breast augmentation, bilateral 0H0V0ZZ Alteration of Bilateral Breast, Open Approach c. Augmentation
Means—method	41 (64.1%)	1 (100%)	37 (59.7%)	8 (44.4%)
Example: Procedure name Code found Missing information	Cryosurgery cervix NMF.GA.AC Destruction of lesion or tissue of cervix Cryosurgery	Arthroplasty and implant elbow with reconstruction fascia lata ligament 239471004 Arthroplasty of the elbow (procedure) Reconstruction method	Cryosurgery cervix 0U5C7ZZ Destruction of Cervix, Via Natural or Artificial Opening Cryosurgery	Arthroplasty and implant elbow with reconstruction fascia lata ligament 1.TM.53.LA-PN Implantation of internal device, elbow joint Reconstruction method
Means—approach	5 (7.8%)		4 (6.5%)	1 (5.6%)
Example: Procedure name Code found Missing information	Radical perineal prostatectomy NGA.JL.AA Radical prostatectomy Perineal approach		Substernal thyroidectomy by cervical approach 0GTK0ZZ Resection of Thyroid Gland, Open Approach Cervical approach	Orchiopexy, inguinal approach 1.QM.74.LA Fixation, testis Inguinal approach
Means—device	3 (4.7%)		1 (1.6%)	2 (11.1%)
Example: Procedure name Code found Missing information	Reduction, primary, open, fracture of femur, open fixation using hip pinning MLB.LD.AA & MLB.DN.AA Open reduction of femur & Implantation of device into bone, femur Hip pin		Reduction intertrochanteric femur fracture, with intramedullary implant with screw and cerclage 0QS[67]04Z Reposition Right/Left Upper Femur with Intramedullary Internal Fixation Device, Open Approach Screw, cerclage	Repair umbilical hernia with prosthesis 1.SY.80.LA Repair, muscles of the chest and abdomen Prosthesis
Total	64 (100%)	1 (100%)	62 (100%)	18 (100%)

CCI: Canadian Classification of Health Interventions; ICD: International Classification of Diseases; ICHI: International Classification of Health Interventions.

Use of ICHI in international statistics

The alignment of the OECD/EU surgical procedure categories and our list of surgical procedures is shown in Table  4. Altogether, 39 of our procedures were encompassed by 19 of the 22 OECD/EU categories. Three of the OECD/EU categories—laparoscopic hysterectomy, laparoscopic repair of inguinal hernia, and stem cell transplantation were not found among our list of procedures. We reviewed the ICHI codes for the 39 procedures to determine whether they would enable the procedures to be accurately assigned to the OECD/EU categories and here are the results:

Table 4.

Procedures encompassed by the OECD/EU surgical procedure categories and their level of representation by ICHI coding

OECD/EU procedure category	Number of procedures in our list and their level of representation in ICHI
	Total number of procedures in category	Full representation without postcoordination	Full representation with postcoordination	Partial representation
Appendectomy	1	1	0	0
Appendectomy, laparoscopic	1	1	0	0
Cataract surgery	2	1	1	0
Cesarean section	3	1	0	2
Cholecystectomy	2	2	0	0
Cholecystectomy, laparoscopic	2	2	0	0
Coronary artery bypass graft	1	1	0	0
Excision of mammary gland, partial	3	3	0	0
Excision of mammary gland, total	3	2	1	0
Hip replacement	2	0	2	0
Hip replacement, secondary	1	0	1a	0
Hysterectomy	6	4	2	0
Hysterectomy, laparoscopic	0	0	0	0
Prostatectomy, open	2	0	0	2
Prostatectomy, transurethral	1	0	1	0
Repair of inguinal hernia	1	1	0	0
Repair of inguinal hernia, laparoscopic	0	0	0	0
Stem cell transplantation	0	0	0	0
Tonsillectomy	2	2	0	0
Total knee replacement	4	0	3	1
Transluminal coronary angioplasty	1	1	0	0
Transplantation of kidney	1	1	0	0
Total	39	23	11	5

EU: European Union; ICHI: International Classification of Health Interventions; OECD: Organisation for Economic Co-operation and Development.

^aNeeds postcoordination for correct OECD/EU category assignment.

1. Procedures fully represented without postcoordination (23 procedures)

   All of them could be assigned correctly to the OECD/EU categories based on the ICHI codes.

2. Procedures fully represented with postcoordination (11 procedures)

   In 10 of the 11 procedures, the information that required postcoordination to capture was not essential in the assignment to the proper OECD/EU categories. For example, in “Mastectomy, bilateral”; “Transurethral prostatectomy, complete” and “Total replacement of hip,” the information highlighted in bold required postcoordination to capture . However, even without postcoordination, these procedures could be assigned to the correct OECD/EU categories based on the ICHI codes. The exception was “Revision of total hip replacement.” Since postcoordination was needed to capture “revision,” this procedure would be assigned to the wrong OECD/EU category Hip replacement without postcoordination, instead of the correct category Hip replacement, secondary.

3. Procedures partially represented (5 procedures)

   All of them could be assigned to the correct category based on their ICHI coding. In all 5 procedures, the missing information (highlighted in bold) did not affect the OECD/EU category assignment (eg, “Radical perineal prostatectomy,” “Low transverse cesarean section,” and “Arthroplasty of knee using cement”).

As long as a procedure is assigned to the correct category, missing information in the ICHI code would not result in over- or undercounting, since the count is based on the number of patients receiving a procedure. Even when multiple codes are used for a procedure in a patient, that procedure is only counted once.

DISCUSSION

Comparison of the procedure coding systems

In this study, we evaluate the ability of the various coding systems to encode surgical procedures as they are recorded in the electronic health record. Since the primary function of the electronic health record is to support patient care, we are examining the fitness of the coding systems for the clinical use case. It is not surprising that SNOMED CT has the best performance in our evaluation, with over 99% full representation. SNOMED CT is primarily a clinical terminology. Most terms in SNOMED CT originate from clinical discourse, which would match well with display names used in the electronic health record. CCI is a close second and can achieve full representation of 92% of procedures. In both SNOMED CT and CCI, most procedures require only a single code. This shows that their granularity is more aligned with clinical documentation compared to ICHI and ICD-10-PCS.

ICHI only achieves full representation in about half of the procedures without postcoordination. This can be partly explained by its relatively small number of codes for surgical procedures (Table  1)—24% of SNOMED CT, 7% of ICD-10-PCS, and 31% of CCI. With postcoordination, full representation in ICHI can be increased to 72%, almost the same as ICD-10-PCS which is 13 times bigger.

ICD-10-PCS is designed primarily for administrative use (eg, reimbursement, casemix). Despite its large number of codes (over 3 times that of SNOMED CT), it only achieves full representation in 72.9% of procedures. One reason for this is the mismatch in granularity requirement between clinical and administrative coding. In some cases, administrative coding requires certain specific detail (eg, for cost determination) which may not be as important in clinical coding. For example, there are 36 ICD-10-PCS codes for “Total hip replacement,” depending on the laterality, type of prosthesis (eg, metal, metal on polyethylene, ceramic, ceramic on polyethylene), and whether bone cement is used. In other cases, different procedures are lumped together in the same ICD-10-PCS code. For example, “Repair of umbilical hernia,” “Repair of ventral hernia,” and “Repair of incisional hernia” are all coded as 0WQF0ZZ Repair Abdominal Wall, Open Approach.

Suggestions to enhance the coverage of ICHI

Given the broad scope of ICHI (ranging from diagnostic tests, medical and surgical procedures to rehabilitation and public health) and its relatively small size, ICHI’s mediocre coverage of surgical procedures is not unexpected. ICHI is designed to support multiple use cases, and statistical reporting and analysis is an important one. As an international classification, it is understandable that ICHI tends to be parsimonious in the creation of stem codes to reduce coding variability and improve comparability among countries. Therefore, improvement in postcoordination capability is probably the best option to expand coverage. Our failure analysis of procedures that cannot be fully represented in ICHI sheds some light on which areas of expansion are more likely to be cost-effective. A lot of missing information falls within the “means” axis, which mostly represents the surgical method, approach, and medical device used in a surgical operation. Compared to “method” and “device,” “approach” is relatively well-covered by existing stem codes. In over 90% of the stem codes, the “means” axis captures the approach. Missing surgical methods account for almost 60% of all partially represented procedures. Among the ICHI extension codes, there is very limited coverage of surgical methods. There is no specific branch of extension codes that covers surgical methods. Only a handful of codes for surgical methods can be found under the branch “Additional descriptive information” (eg, tissue flaps or grafts, homograft). As a result, most missing surgical methods cannot be captured with postcoordination. Adding some common surgical methods (eg, laser, cryosurgery, ligation) to the extension will help to improve ICHI’s coverage of surgical procedures.

While general types of surgical approach (eg, open, percutaneous endoscopic approach) are well-covered by existing stem codes, some approaches for specific body locations (eg, retropubic, perineal, cervical approach) are missing. Sometimes, information about the approach can have clinical importance (eg, risks and types of complication may differ with approach). ICHI should consider adding approaches for specific body locations as extension codes. “Target” is another common type of missing information, accounting for almost a third of cases. What is usually required is a body structure more specific than that captured in the “target” axis. To this end, ICHI already has 3025 extension codes under “Specific Anatomic Detail.” However, it is surprising that common anatomic entities such as “ulnar nerve,” “neck of femur,” “anterior vaginal wall” are still missing. A systematic review of the anatomic codes in the ICHI extension seems warranted.

For the use case of collecting international statistics on surgical procedures, ICHI seems to be generally adequate. Most of the procedures in our study can be assigned to the correct OECD/EU category based on their ICHI encoding. One exception is that the procedure “Revision of total hip replacement” would be wrongly classified without postcoordination. Postcoordination is a brand-new feature in the WHO-FIC classification systems (ICD-11 also supports postcoordination) and will have impact on tooling, coder education, and coding variability. Since postcoordination may not be universally implemented in all WHO member countries, ICHI should consider adding a new stem code for “revision hip replacement” to ensure that the capture of OECD/EU statistics can still be supported without postcoordination.

Limitations and future work

We recognize the following limitations in this study. The scope of ICHI is very broad and covers many different types of health interventions. Our study only concentrates on surgical procedures. The list of commonly performed surgical procedures is derived from one large US healthcare provider and may not be representative of other healthcare settings. The coding of the procedures in each coding system was done independently by 2 terminologists but was not externally validated. In the future, we plan to evaluate ICHI for other use cases (eg, casemix, quality measurement) and how ICHI can be mapped to other procedure coding systems. Study in these areas would contribute to ICHI’s future development and improvement.

CONCLUSION

For 229 commonly performed surgical procedures, ICHI was able to fully represent 52.8% without postcoordination. A further 19.2% could be fully represented with postcoordination. For the 27.9% that could only achieve partial representation, missing method (58.6%) was the most common reason for the failure of full representation, followed by missing target (30%). For the clinical use case, the performance of ICHI is inferior to clinical coding systems (eg, SNOMED CT) but is comparable to ICD-10-PCS. There is room for improvement of coverage in ICHI by enhancing its postcoordination capability. For the statistical use case, ICHI coding is generally adequate and could support correct assignment to the appropriate OECD/EU categories.

FUNDING

This research was supported in part by the Intramural Research Program of the NIH, National Library of Medicine.

AUTHOR CONTRIBUTIONS

KWF, JX, and FA conceived and designed the study. JX, FA, LB, and JM performed the coding in the various coding systems. KWF performed the primary data analysis. KWF drafted the manuscript and all authors contributed substantially to its revision.