Population-Based and Routinely Collected Health Data Studies of Immunobullous Diseases: A Scoping Review

Abstract

Background:

Validated case definitions are essential for reliable epidemiological research using routinely collected health data. The extent to which population-based studies on autoimmune blistering diseases (AIBDs) use validated case definitions is uncertain, and variability in definitions may limit the generalizability of findings.

Objective:

To describe definitions of AIBDs used in population-based epidemiologic studies and whether these definitions have been validated.

Methods:

MEDLINE and Embase were searched for epidemiologic studies on pemphigoid, pemphigus and their variants, dermatitis herpetiformis (DH), epidermolysis bullosa acquisita (EBA), and linear IgA bullous dermatosis (LABD). Eligible studies reported case definitions using routinely collected data. Two reviewers independently screened studies and extracted data, with discrepancies resolved by consensus.

Results:

Ninety-two studies met the inclusion criteria. Among 63 pemphigoid studies, the majority used International Classification of Diseases (ICD) codes, and 26 studies validated case definitions using combinations of clinical diagnosis, histopathology, and direct or indirect immunofluorescence, often with scoring systems. The majority of 39 pemphigus studies used ICD codes or electronic medical records to identify cases; of the 5 studies that validated cases, most used chart review requiring 2 or more confirmatory clinical or laboratory findings. Two of 6 DH studies used validated definitions with clinical and immunopathological criteria. Among 8 EBA studies, none reported formal validation. One LABD study was included without case validation.

Conclusion:

Most population-based studies of AIBDs rely on administrative coding or electronic medical records, but few use formally validated case definitions. Improving and standardizing validation practices is important to enhance the accuracy and comparability of epidemiological findings in AIBDs.

Introduction

Validated case definitions are essential for reliable epidemiological research using population-based and routinely collected health data.^1-5 Autoimmune blistering diseases (AIBDs) present unique challenges for case ascertainment because diagnosis often relies on integrating clinical features with histopathological and immunopathological findings. The diagnostic gold standard is direct immunofluorescence (DIF) microscopy of perilesional biopsy specimens, often complemented by histology,^6,7 yet such confirmatory data are rarely available in large administrative databases. ⁸ Misclassification can occur when diagnostic codes are applied inconsistently, when there is overlap with other conditions, or when classification is based on incomplete clinical information.

While population-based studies offer valuable opportunities to study the epidemiology and outcomes of AIBDs, their validity depends on the accuracy of the algorithms used to identify cases. In other disease areas, validated definitions have been shown to improve case ascertainment and reduce bias,^5,9,10 but for AIBDs, there is limited understanding of how case definitions are created and applied and whether they have undergone formal validation. Many existing studies in AIBDs do not report validation processes,^11-14 making it difficult to interpret and compare results across jurisdictions.

The objective of this scoping review is to systematically identify population-based and routinely collected health data studies focusing on AIBDs, describe the case definitions used, and determine whether these definitions have been validated.

Materials and Methods

We searched for epidemiologic studies on AIBDs, including bullous pemphigoid (BP), mucous membrane pemphigoid (MMP), pemphigus vulgaris (PV), pemphigus foliaceus (PF), dermatitis herpetiformis (DH), epidermolysis bullosa acquisita (EBA), and linear IgA bullous dermatosis (LABD).

Search Strategy

We searched MEDLINE and Embase on April 8, 2025. Variations of the search concepts for the 5 groups of AIBD of interest were applied: “bullous pemphigoid” or “mucous membrane pemphigoid,” “pemphigus vulgaris” or “pemphigus foliaceous,” “linear IgA bullous disease,” “dermatitis herpetiformis,” and “epidermolysis bullous aquisita” (Supplementary Table 1). The search strategy was developed with assistance from an information specialist. Searched articles were exported to Covidence (Melbourne, VIC, Australia; www.covidence.org) for review.

Inclusion and Exclusion Criteria

Our inclusion criteria required each study to (1) be an epidemiologic study using routinely collected health data reporting 1 or more of the 5 groups of AIBD as an exposure or outcome, and (2) describe the disease definition of each AIBD used in the study. Study types included population-based, prospective, and retrospective cohort, case-control, and cross-sectional designs. The reference lists of relevant systematic reviews were additionally screened to identify further studies for inclusion. We excluded case series with fewer than 100 patients with the disease of interest, studies that focused on the treatment of AIBDs, gray literature, abstracts, conference proceedings, unpublished studies, and ongoing studies. We did not restrict by geographic, language, patient age, or date.

Data Extraction

Two reviewers independently screened titles and abstracts, and full length of papers, in duplicate (H.J.Z., V.W., M.G., I.U.). For the first 5 studies, we piloted the data extraction table among all reviewers, followed by a consensus meeting to ensure harmonization. This process was repeated twice until all authors’ initial evaluations reached >80% agreement. For subsequent studies, results for the extraction table from 2 independent reviewers were compared, and discrepancies were resolved through discussion until consensus was reached. If discrepancies could not be resolved, a third author was consulted. When studies reported more than 1 AIBDs, data were extracted separately for each disease, as case definitions and validation methods often differed between conditions (e.g., pemphigus vs pemphigoid).

We used Preferred Reporting Items for Systematic Reviews and Meta-analyses extension for Scoping Review to guide reporting. ¹⁵

Results

The search yielded 2944 studies after duplicates were removed. After title and abstract screening, 260 studies underwent full-text review. A total of 92 full-text records were included, including 13 additional studies identified from reference lists of systematic reviews (Supplementary Figure 1).

Pemphigoid (BP and MMP)

Sixty-three studies reported on pemphigoid (Supplementary Table 2), with 49 studies evaluating BP alone, 6 included both BP and MMP, and 8 reported pemphigoid without subtype specification. Forty-six studies using cases that were ascertained through varying International Classification of Diseases (ICD) coding systems (ICD-8 694.5, 694.6; ICD-9 694.5, 694.9; ICD-10 L12.0, L12.8, L12.9).^11,14,16-59 Three studies used Systematized Nomenclature of Medicine-Clinical Terms (SNOMED) code,^60-62 which is a U.S. standardized system for recording clinical information in electronic health records. ⁶³ Four studies used Read codes,^64-67 which are standardized codes used in the United Kingdom (UK) patients’ primary care records. ⁶⁸ Supplementary Table 2 lists the specific SNOMED and Read codes for pemphigoid. Ten studies obtained pemphigoid cases using electronic medical records, with some requiring an additional combination of clinical features, such as histopathology, DIF, and/or indirect immunofluorescence (IIF), and enzyme-linked immunosorbent assay (ELISA), and most studies required at least 2 to 4 of these criteria for case confirmation.^12,69-77

A total of 26 studies had validated pemphigoid case definitions.^{18,19,21,30,32,38-40,42,43,45,47,50-54,56,57,59,62,65-67,69,77} Eight Taiwanese studies^{18,21,42,51,53,54,56,59} used the same validation source, reporting a positive predictive value (PPV) of 98% (95% CI 96.3%-99.7%) based on patients’ clinical information, histopathology, immunofluorescence, and medical records. ⁵³ Another study validated BP case definitions using a scoring system (clinical presentation, histopathology, DIF, or IIF). It reported varying PPVs (48.4%-71.8%), negative predictive value (NPV; 83.1%-91.5%), sensitivity (86.3%-89.4%), and specificity (55.7%-67.1%) for cases defined by singular diagnostic tools (histology, DIF, or IIF). ¹⁹ Two Finnish studies^39,40 cited the same study that validated BP case definition by manually reviewing at least 3 registrations of ICD-10 code in a patient’s record, which reported a PPV of 73.6% (95% CI 71.0%-76.0%), sensitivity of 85.1% (95% CI 83.1%-87.1%), and specificity of 58.3 (95% CI 55.6%-61.1%). ³⁹ Two Swedish studies^30,52 cited the same study that validated BP cases by cross-checking ICD-10 code with SNOMED code and manual chart review of histology, DIF, or ELISA results, reporting a PPV of 92%. ³⁰ Four studies^32,38,57,62 cited the same validation study that used a manual scoring system with at least 2 points for each of positive clinical diagnosis, histopathology, DIF, and IIF/ELISA. The study reported a high PPV (99%, 95% CI 93%-99%) and specificity (99%, 95% CI 97%-99%), but low sensitivity (37%, 95% CI 31%-44%), and NPV (38%, 95% CI 33%-43%). ³² Three UK studies^{65 -67} cited the same article that validated cases by manually cross-checking Read codes with inpatient ICD-10 codes, which reported a PPV of 93.2% (95% CI 91.3%-94.8%). ⁶⁷ Six studies described validating their cases but did not report or cite numerical validation measures.^{38,43,47,50,69,77}

Pemphigus (PF and PV)

In 39 pemphigus studies (Supplementary Table 3), 3 studies evaluated PV alone, 9 included both PV and PF, and 27 reported pemphigus without subtype specification. Twenty-four used ICD codes (ICD-8 684, ICD-9 694.4, ICD-10 L10.0, L10.1, L10.2, L10.4, L10.9),^{11,14,16,17,21,25,26,32 -34,46,48,51,57,62,78 -85} 2 of which required the code in at least 2 outpatient dermatologist reports or once in an inpatient discharge record.^51,78 One study used SNOMED codes, ⁸⁶ 1 study used Read codes, ⁶⁷ and 13 studies identified cases via electronic medical records, with or without multiple physician entries for pemphigus, hospital discharge pemphigus diagnoses, or immunofluorescence results.^12,75,87-97 Supplementary Table 3 lists the specific SNOMED and Read codes for pemphigus. An English study validated PV cases by manually cross-checking Read codes with inpatient ICD-10 codes, reporting a PPV of 58.5% (95% CI 48.0%-68.9%). ⁶⁷ Four studies^32,57,62,80 cited the same validation study that used manual chart review and a scoring system of meeting at least 2 of the clinical diagnosis, histopathology, DIF, and IIF/ELISA. The study reported high PPV (100%, 95% CI 96%-100%) and specificity (100%, 95% CI 98%-100%), but low sensitivity (47%, 95% CI 40%-54%) and NPV (35%, 95% CI 30%-41%). ³²

Dermatitis Herpetiformis

Six population-based studies on DH were identified (Supplementary Table 4). Case identification methods varied: 3 studies used ICD codes with varying versions (ICD-7 70 400, ICD-8 69 399, ICD-10 L13.0),^11,98,99 2 studies used electronic health records,^100,101 and 1 study reported using Read codes, although specific codes were not provided. ¹⁰² Of these 6 studies, 2 validated their case definitions. Albadri et al reported a PPV of 62.5% from manually reviewing 160 cases for positive DIF or at least 2 of the following: clinical presentation, serology, histopathology, response to dapsone, or duodenal biopsy. ⁹⁸ Sigurgeirsson et al manually reviewed a subset of 63 cases from 976 patients, based on clinical and histopathological findings, and reported a PPV of 98%. ⁹⁹

Epidermolysis Bullosa Acquisita

Seven of 8 studies with EBA cases used the ICD-10 code (L12.3), with sample sizes ranging from 2 to 1344 patients (Supplementary Table 5).^{14,26,33,34,37,48,103} The multicentre study spanning Germany, Italy, Singapore, Israel, and the Netherlands used electronic medical records containing clinical and immunofluorescence results to identify EBA cases. ¹² No formal validation of the EBA case definition was reported across all studies.

Linear IgA Bullous Dermatosis

For LABD, one recently published study was identified, but did not use a validated definition (Supplementary Table 6). The study used the Korean National Health Insurance Database to identify patients who had received dapsone treatment within 6 months of a skin biopsy, based on unspecified drug and procedure codes. ¹⁰⁴

Discussion

Most population-based and routinely collected health data studies on AIBDs rely on administrative codes (ICD, SNOMED, Read) or electronic medical records, with few studies incorporating confirmatory clinical, histopathological, and/or immunopathological data. Pemphigoid studies demonstrated the highest proportion of validated case definitions, often using scoring systems combining clinical diagnosis, histopathology, DIF, IIF, and ELISA. In contrast, studies on rarer diseases such as DH, EBA, LABD, and pemphigus rarely use validated definitions.

Validation methods varied widely in pemphigoid studies. Several validated studies cited the same validation article, which reduced opportunities to assess validation methodologies. For example, 8 of 26 validated pemphigoid studies^{18,21,42,51,53,54,56,59} cited Wu et al but the authors did not specify weights or decision rules. ⁵³ This lack of transparency limits reproducibility and comparability. Another study reported single-test PPVs for pemphigoid: histopathology 68.4%, DIF 71.6%, and IF 48.4%, ¹⁹ demonstrating reliance on individual tests reduces accuracy. Together, these findings illustrate the heterogeneity of existing validation methods and underscore the need for standardized, multi-criteria algorithms for AIBDs. Broader adoption of validated definitions would improve accuracy, comparability, and support multinational research. Comparable advances in cardiology and chronic disease epidemiology showed that standardized validation using ICD-10 algorithms and electronic health record based approaches reduces misclassification and strengthens population-level estimates.^9,10,105 A similar approach in dermatology could improve case ascertainment, harmonize epidemiological estimates across regions, and inform clinical guidelines and resource planning.

Although no EBA or LABD cases were validated, DH and pemphigus studies reported a wide range of PPVs, ranging, respectively, from 56.25% to 98% and 50.8% to 100%.^{32,57,62,67,80,98,99} These findings indicate the validity of the case definition is uncertain and highly variable. Furthermore, 1 study showed that a high PPV can be accompanied by low sensitivity and NPV. ³² Reporting PPV alone overlooks the importance of sensitivity and NPV, such as omitting true cases or failing to exclude non-cases. While such algorithms may be suitable for outcome studies requiring diagnostic certainty, they are less reliable for estimating incidence or prevalence in rare diseases such as AIBDs. Low sensitivity underestimates disease frequency by missing true cases. Low NPV weakens confidence in ruling out disease, which is crucial for understanding the clinical utility of the algorithm and the overall disease burden. ¹⁰⁶ Complete reporting of validity measures, sensitivity, specificity, PPV, and NPV, is therefore important in future validation frameworks, with reporting guidelines already encouraging the report of all validation measures. ¹⁰⁷

Strengths of this review include systematically mapping population-based and routinely collected AIBD studies, identifying both validated and unvalidated approaches. A comprehensive search strategy across 2 major databases, inclusion of multiple disease subtypes, and duplicate screening and data extraction enhance rigor and reproducibility. By separating findings by 5 disease subtypes, this review also provides condition-specific insights that can inform targeted methodological improvements.

Geographic data concentration, particularly the predominance of validated cohorts from Taiwan, limits generalizability to other health systems. Small sample sizes in rarer conditions such as DH, EBA, and LABD further constrain conclusions. BP and MMP were grouped because of the small number of studies reporting MMP and the inclusion of studies that did not specify pemphigoid subtypes. Similarly, PV and PF were grouped because a few studies reported specific pemphigus subtypes. These groupings are a limitation, as BP and MMP differ in epidemiology, etiology, and clinical outcomes, as do PV and PF. Finally, although the search was comprehensive, some relevant studies may have been missed.

Conclusion

AIBD case validation is an important consideration in studies using routinely collected data, and studies to date have been highly variable in their use of validated case ascertainment. Future large database studies on AIBDs would be strengthened by using validated case definitions, enabling their findings to be implemented with higher certainty in the clinic and for public health.

Supplemental Material

sj-docx-1-cms-10.1177_12034754261442933 – Supplemental material for Population-Based and Routinely Collected Health Data Studies of Immunobullous Diseases: A Scoping Review

Supplemental material, sj-docx-1-cms-10.1177_12034754261442933 for Population-Based and Routinely Collected Health Data Studies of Immunobullous Diseases: A Scoping Review by Heather J. Zhao, Vince Wu, Michelle Gandelman, Inna Ushcatz, Anastasiya Muntyanu and Aaron M. Drucker in Journal of Cutaneous Medicine and Surgery