Patient diagnoses must be classified according to definitions that are consistent across countries and over time if health records are to be used for research in epidemiology or the management of health care systems. Consistency is provided by applying diagnostic coding systems, the most widely used of which is the one maintained by the World Health Organization (WHO): the International Classification of Diseases, currently in its 10th revision (the ICD-10).1 This revision of the ICD was completed in 1992 and its latest version was published in 2016. The ICD-10 is a simple alphanumeric coding system that facilitates diagnostic classification for both clinical and epidemiologic purposes. It provides the data for computing mortality and morbidity rates in countries belonging to the WHO, and some countries use it to allocate health care resources.2

Certain ICD-10 codes have been reported to be imprecise.3 Recent studies have found deficiencies in the classification of allergic conditions2,4: as a result, events such as anaphylaxis are rendered less visible by underreporting, data collection is compromised, and statistical analyses are less precise, possibly leading to underfunding. A new revision —the ICD-11 —is currently being developed with the aim of solving some of these problems.1

An additional concern is that the ICD-10 coding of dermatologic diseases includes areas of difficulty and imprecision, the reason why an adaptation for this specialty was developed in Spain in 1999.5 The adapted version was developed to complement, not replace, the general ICD.

The literature on the shortcomings of the ICD-10 is scant, and an analysis of the reasons for coding problems in dermatology could provide useful information and point toward modifications that could be incorporated into the draft ICD-11. Such an analysis could also provide verification that the drafted revisions are going in the right direction.

The recent DIADERM study included nearly 11000 diagnoses representative of routine clinical practice in outpatient dermatology clinics in Spain.6 Data for 8832 patients were contributed by a stratified random sample of 80 Spanish dermatologists. Diagnoses were collected on 3 consecutive days in January and 3 consecutive days in May, and the diagnoses were coded by a dermatologist familiar with the ICD-10 adapted for dermatology.5 The present study aims to identify all the cases in the DIADERM database that were not assigned a code, analyze the reasons for difficulty, identify problems that could be eliminated through revision, and determine whether the draft ICD-11 has improved classification by correcting the shortcomings of the ICD-10.

We found 101 cases with diagnoses that had not been assigned a code. After excluding unintelligible or unclear notes and nondermatologic diseases, there remained 41 dermatology cases to analyze (Fig. 1).

Figure 1.

Flow chart of cases whose diagnoses were included for analysis in this study.

(0.05MB).

The DIADERM study allowed us to obtain data for analyzing problems of diagnostic classification in routine dermatologic practice in Spain. Our analysis leads us to conclude that both the ICD-10 adapted for Spanish dermatology5 and the draft ICD-11 will correctly classify the vast majority of clinical diagnoses we make: only 0.20% of the diagnoses in the DIADERM database corresponded to conditions that were absent from the adapted ICD-10 and that percentage fell to 0.15% when we used the ICD-11.

Some of the most important dermatology classification problems of the ICD-10 had already been reported and corrected in the version adapted for dermatology in Spain.5 Nevertheless, this adaptation still contains shortcomings that affect the coding of certain diagnoses. A systematic review of all the classifications lies outside the aims of this study, and given our finding that only a small number of dermatologic diagnoses lack codes, it is difficult to see which parts of the system work better or worse. However, we have observed problems in the following areas: 1) defining the exact topography of certain diseases in which that information may be of interest (in genital eczema and metastatic Crohn disease, for example) and 2) the lack of inclusion of certain diseases, possibly because they have been described or developed only in recent decades. Examples are frontal fibrosing alopecia9; vascular malformations, whose classification and nomenclature have changed10; progressive macular hypomelanosis11; pigmentary mosaicism12; indolent acral CD8+ lymphoma13; amicrobial flexural pustulosis14; infantile perianal protrusion15; and aquagenic keratoderma.16 Sebopsoriasis also lacks a specific code, and myxoid pseudocyst seems to fall into the same category as cutaneous myxoma (as a mucous cyst) — an unfortunate decision given that a myxoid pseudocyst has no epithelial lining.17

Improvements have been made in the draft ICD-11, as our report reveals. The new system allows the user to combine a root code with one or more extension codes. This flexibility helps the coder define the exact location of a symptom or the area affected by a disease.8 The new revision also includes specific codes for several diseases not included in the ICD-10 adapted for dermatology in Spain5; in fact the new classification of vascular anomalies is a substantial improvement.8 Most of the remaining terms that lack a specific code are in the draft ICD-11 index, where they have higher-level codes indicated for them. It is true that there are 3 diagnoses in the Spanish adaptation of the ICD-10 that still lack codes in the draft ICD-11; however, all of them are present in the index. It is reasonable for a general classification of diseases, such as the ICD, to include a certain degree of “granularity” in the interest of reducing the number of codes.1,5 Here is where adapting the system to the needs of specialties can play a part in achieving greater detail,5 a feature that is essential for compiling statistics on disease prevalence and ensuring that underreporting does not take place.2,4

Developing a new classification system requires the collaboration of experts in the material covered, and the Dermatology Topic Advisory Group is currently working on the draft ICD-11.18 Studies to detect coding problems are particularly important at this time if we are to develop new categories. One study that detected problems in the coding of anaphylaxis,4 for example, led to expert collaboration and new codes for the draft ICD-11.19 We believe our study provides a basis for improving the classification of dermatologic diseases. In our view, all the diagnoses mentioned in our Table 3 should receive specific codes in a future dermatology adaption of the ICD-11, and those entities should at least be included as indexed terms in the general version of the tool, as their existence is well established in the literature.9–16 We also believe that frontal fibrosing alopecia should be assigned its own code. Even though incidence and prevalence data are lacking for this condition, it is not an uncommon diagnosis, in our experience, and that of others.20

We think that most of the DIADERM diagnoses that lack codes in the ICD-10 adapted for Spanish dermatology5 have been included in this study, given that the error rate of the initial coder was low and 3 dermatologists revised any coding for which doubts arose.6 However, it is still possible that we may have failed to include some diagnoses that had been assigned to higher-level categories in the database. Another important limitation is that the review of uncoded diagnoses in the present study — and thus their inclusion or exclusion as diagnoses that lacked a specific code in the ICD-10 for dermatology or the draft ICD-11 — was managed by a single coder (G. G. L.), in spite of the admitted degree of subjectivity involved. We think these limitations do not substantially change the conclusions that can be reached.

In summary, we conclude that the ICD-10 adapted for dermatology in Spain5 and the draft ICD-11 appropriately classify the vast majority of diagnoses made routinely in dermatology practices. Since the methods used to create the DIADERM database ensure that the diagnoses listed are highly representative of practice in this country,6 we conclude that these classification systems allow us to study the real characteristics of our caseloads. Nonetheless, we were able to detect shortcomings in the dermatology adaptation of the ICD-10,5 particularly with respect to diagnoses that have emerged in recent decades. Classification systems clearly need to be kept up to date, and once the final version of the ICD-11 is published it would be useful to form a group to collaborate on that task. We think that including the suggestions we have made for the ICD-11, and also adapting the new system for dermatology, would help with the problem of underreporting as well as with the compiling of statistics on mortality rates in dermatology and the prevalence of currently underdiagnosed skin diseases.

The DIADERM study, which is supported by the Healthy Skin Foundation of the Spanish Academy of Dermatology and Venereology (AEDV), has received funding from Novartis. Novartis did not participate in data collection, data analysis, or interpretation of the results. Guillermo González-López received a Juan de Azúa Grant from the AEDV's Healthy Skin Foundation to support his work on this paper.

The authors do not recognize any conflicts of interest in relation to this article.