Abstract
BACKGROUND: Muscular dystrophies (MDs) comprise a heterogenous group of genetically inherited conditions characterized by progressive muscle weakness and increasing disability. The lack of separate diagnosis codes for Duchenne MD (DMD) and Becker MD, 2 of the most common forms of MD, has limited the conduct of DMD-specific real-world studies.
OBJECTIVE: To develop and validate administrative claims–based algorithms for identifying patients with DMD and capturing their nonambulatory and ventilation-dependent status.
METHODS: This was a retrospective cohort study using the statistically deidentified Optum Market Clarity Database (including patient claims linked with electronic health records [EHRs] data) to develop and validate the following algorithms: DMD diagnosis, nonambulatory status, and ventilation-dependent status. The initial study sample consisted of US patients in the database who had a diagnosis code for Duchenne/Becker MD (DBMD) between October 1, 2018, and September 30, 2020, who were male, aged 40 years or younger on their first DBMD diagnosis, and met continuous enrollment and 1-day minimal clinical activities requirement in a 12-month measurement period between October 1, 2017, and September 30, 2020. The algorithms, developed by a cross-functional team of DMD specialists (including patient advocates), were based on administrative claims data with International Classification of Diseases, Tenth Revision, Clinical Modifications coding, using information of diagnosis codes for DBMD, sex, age, treatment, and disease severity (eg, evidence of ambulation assistance/support and/or evidence of ventilation support or dependence). Patients who met each algorithm and had EHR notes available were then validated against structured fields and unstructured provider notes from their own linked EHR to confirm patients’ DMD diagnoses, nonambulatory status, and ventilation-dependent status. Algorithm performance was assessed by positive predictive value with 95% CIs.
RESULTS: A total of 1,300 patients were included in the initial study sample. Of these, EHR were available and reviewed for 303 patients. The mean age of the 303 patients was 14.8 years, with 61.7% being non-Hispanic White. A majority had a Charlson comorbidity index score of 0 (59.4%) or 1-2 (27.7%). Positive predictive value (95% CI) was 91.6% (85.8%-95.6%) for the DMD diagnosis algorithm, 88.4% (80.2%-94.1%) for the nonambulatory status algorithm, and 77.8% (62.9%-88.8%) for the ventilation-dependent status algorithm.
CONCLUSIONS: This work provides the means to more accurately identify patients with DMD from administrative claims data without a specific diagnosis code. The algorithms validated in this study can be applied to assess treatment effectiveness and other outcomes among patients with DMD treated in clinical practice.
DISCLOSURES: This study was funded by Pfizer, which contracted with Optum to perform the study and provide medical writing assistance. Ms Schrader reports being an employee of Parent Project Muscular Dystrophy. Mr Posner reports being an employee and stockholder of Pfizer and receiving support from Pfizer for attending conferences not related to this manuscript. Dr Dorling reports being an employee and stockholder of Pfizer at the time the study was conducted and is a current employee of Chiesi USA, Inc. Ms Senerchia reports being an employee of Optum and owning stock in Pfizer and UnitedHealth Group, the parent company of Optum. Dr Chen reports being an employee and stockholder of Pfizer. Ms Beaverson reports being an employee of Pfizer and owning stock in Pfizer and Amicus Therapeutics. Dr Seare reports being an employee of Optum at the time the study was conducted. Dr Garnier and Ms Merla report being employees of Pfizer. Ms Walker reports being an employee of Optum. Dr Alvir reports being an employee and stockholder of Pfizer. Dr Mahn reports being an employee and stockholder of Pfizer. Dr Zhang reports being an employee of Optum. Ms Landis reports being an employee of Optum. Ms Buikema reports being an employee of Optum and holding stock in UnitedHealth Group, the parent company of Optum.
Plain language summary
We developed a method to identify patients with Duchenne muscular dystrophy (DMD) from health insurance claims. Using electronic health records to confirm diagnoses, we found the method correctly identified 92% of patients with DMD. We developed similar methods to determine whether patients with DMD were unable to walk (88% accuracy) or used machines to help with breathing (78% accuracy). These methods can be used in studies of DMD treatment effectiveness outside of clinical trials.
Implications for managed care pharmacy
We developed an administrative claims–based algorithm with a positive predictive value (PPV) of 92% for identifying patients with DMD, as well as algorithms for accurately determining nonambulatory status (PPV = 88%) and ventilation dependence (PPV = 78%) in this patient population. These algorithms can facilitate the conduct of observational studies to assess treatment patterns and outcomes among patients with DMD treated in clinical practice.
Muscular dystrophies (MDs) comprise a heterogenous group of genetically inherited conditions characterized by progressive muscle weakness and increasing disability.1 The prognosis for patients with MD varies according to the types of gene and muscle involved, the age of onset and typical age at diagnosis, and the severity of disease.1
The most common subtype of MD is Duchenne MD (DMD), which has an estimated prevalence of 10.2-12.6 per 100,000 male patients.2,3 DMD is caused by loss-of-function mutations in the X-linked gene for dystrophin, a protein required for normal muscle structure and function.1 Lack of dystrophin expression causes progressive loss of muscle strength, leading over time to loss of ambulation and upper limb function as well as cardiac and respiratory complications that may require nocturnal and/or daytime ventilation use.4 DMD is the most severe dystrophinopathy, featuring early-childhood onset (symptoms appearing around age 2 or 3); rapid progression; and severe complications that typically result in loss of ambulation by age 15, ventilation dependence by the second decade of life, and death between the ages of 20 and 40.5-7 A less severe form of dystrophinopathy, Becker MD (BMD), has a lower estimated prevalence of 1.5-3.6 per 100,000 male patients2,3 and a more variable clinical presentation. Symptom onset in BMD typically occurs after 5 years of age and into adulthood, with patients potentially remaining ambulatory for decades and having substantially lower risk of early mortality than those with DMD.8
The current standard of care for DMD comprises supportive care adapted to patients’ disease progression, including nutritional management, ventilation and other respiratory care, and treatment of cardiomyopathy as needed.9,10 The majority of patients with DMD are also treated with corticosteroids, which have been shown to improve muscle function, delay loss of ambulation, and increase life expectancy in this population.11-13 However, corticosteroids are associated with substantial endocrine and metabolic side effects and do not cure the underlying disease.8,14,15 In addition, 4 antisense oligonucleotides are approved/conditionally approved by the US Food and Drug Administration for use in patients with DMD who have certain deletions amenable to exon skipping.16 These therapies use splice switching to selectively skip DMD exons adjacent to mutations creating partially functional dystrophin proteins. Additionally, there are numerous clinical trials ongoing to evaluate therapies designed to restore dystrophin, as well as drugs targeting the secondary pathologies associated with dystrophin loss, such as inflammation, fibrosis, mitochondrial dysfunction, and impaired muscle regeneration.9,17 Although clinical trials are the gold standard for evaluating treatment efficacy and safety,18 the recruitment and retention of sufficient patients to conduct trials designed to examine long-term outcomes is challenging because of the rarity, heterogeneity, and progressive nature of DMD. In addition, results from clinical trials, which have rigorously enforced protocols and highly selected study participants, may not always reflect outcomes among patients receiving routine care in clinical practice.19
One approach to measuring routine care and outcomes, such as cost, disease burden, treatment patterns, and natural history, in MD are study designs that use administrative claims data to conduct real-world observational studies.20-25 But identifying patients from claims comes with its own challenges, as the use of administrative diagnosis codes that were not developed for research purposes can result in some misclassification when relied upon to identify disease. This limitation may be more impactful in studies of rare conditions, such as MDs, which often lack unique diagnosis codes.26 There are codes for general MD, and in October 2018, a new International Classification of Diseases, Tenth Revision, Clinical Modifications (ICD-10-CM) diagnosis code for Duchenne/Becker MD (DBMD), G71.01, was introduced.27 Although this was an improvement over previous iterations of the ICD in which all forms of MD were conflated, it remained impossible to distinguish patients with DMD from those with BMD solely on the basis of a diagnosis code.
Thayer et al developed a claims-based algorithm to identify probable DMD in a sample of patients with at least 1 code for general MD: ICD-9-CM 359.0 (congenital hereditary MD) or ICD-9-CM 359.1 (hereditary progressive MD) between July 1, 2000, and December 31, 2009.22 The algorithm contained a large list of diagnostic codes, pharmacy prescriptions, and procedure codes unique to DMD management and consistent with clinical milestones typical for DMD; however, the study did not include validation. Soslow et al used claims data from January 1, 2003, to September 30, 2015, and validated their algorithm with local medical records on a subset of patients.23 The algorithm improved the identification of DMD or BMD from 55% using the general diagnosis code for all MDs (ICD-9 359.1) alone to 77% using an algorithm that included the code in addition to specific exclusion criteria. Chen et al evaluated the performance of a simpler administrative claims algorithm developed by Parent Project Muscular Dystrophy, which included the criteria of at least 2 office visits with a diagnosis of MD using the ICD-9-CM general diagnosis code for all MDs (ICD-9-CM 359.1) and the ICD-10-CM code (ICD-10-CM G71.0: MD, unspecified), as well as male sex and age of 18 years or younger at the first MD diagnosis.24 Validation of the algorithm against medical chart review using data from January 1, 2013, to December 31, 2016, yielded positive predictive values (PPVs) of 95% for MD overall and 85% for DMD or BMD, and 74% for DMD, respectively.24
After several years of data, including the new ICD-10-CM code G71.01 for DBMD, had accumulated, the goal of the present study was to build upon the prior work by Chen et al by developing and validating a simple claims-based algorithm to identify patients with DMD based on the new more specific code G71.01. Algorithms to capture nonambulatory and ventilation-dependent status were also developed. All algorithms were validated using data from electronic health records (EHRs).
Methods
STUDY DESIGN AND DATA SOURCE
This was a retrospective observational study using administrative claims linked to deidentified EHR data. Review of structured fields and unstructured provider notes from EHR were used to validate the performance of claims-based algorithms that were developed for identifying patients with DMD and determining whether patients were nonambulatory or ventilation dependent.
Algorithms were developed by a cross-functional team of DMD clinical, coding, and therapeutic area specialists, including patient advocates from Parent Project Muscular Dystrophy. On the basis of a targeted literature review, initial algorithms based on information regarding the ICD-10 diagnosis code of DBMD (G71.01), sex (male), age (≤40 years), treatment (ie, prescription of glucocorticoids or exon-skipping therapy), and disease severity (ie, evidence of ambulation assistance/support, evidence of ventilation support/dependence) were proposed for each algorithm. Diagnosis and procedure codes suggesting ambulation assistance/support or ventilation support/dependence and procedure or national drug codes identifying specific DMD treatments were then identified by clinician-guided coding specialists.
Data used for algorithm development and validation were from the Optum Market Clarity Database, a proprietary database that links administrative claims data from Optum-affiliated and nonaffiliated payers in the United States with EHR (including structured fields and unstructured provider notes) from Optum’s EHR database. The statistically deidentified integrated database comprises a comprehensive collection of demographic, clinical, operational, and financial information from inpatient and outpatient encounters for more than 75 million unique patients, including but not limited to vital signs and other observable measures, laboratory results, outpatient prescriptions written, inpatient medications administered, and coded diagnoses and procedures. Notes that were linked to the Optum Market Clarity Database were used to maximize the number of notes available for validation. The earliest note was dated March 10, 2003, through September 30, 2020. During the patient identification period, more than 25 million patients with health plan eligibility had both clinical and claims data in the database.
The study protocol was reviewed by the WCG Institutional Review Board (June 15, 2021), which waived the requirement for informed consent for use of protected health information because the risk to individuals’ privacy was determined to be minimal.
STUDY POPULATION
The study population included US patients in the Optum Market Clarity Database who were enrolled in commercial, Medicaid, or Medicare Advantage health plans and had a diagnosis of DBMD (ICD-10-CM G71.01) in claims between October 1, 2018, and September 30, 2020 (patient identification period) (Figure 1). Patients were also required to have continuous health plan enrollment in claims with medical coverage for at least 1 year between October 1, 2017, and September 30, 2020 (claims observation period) (Figure 1), and to have clinical activity in the EHR database that overlapped the claims observation period for at least 1 day. A 12-month measurement period was created within the claims observation period for each patient and was assigned on the basis of overlapping periods of continuous enrollment in the claims with the first DBMD diagnosis date and presence of provider notes in the EHR, maximizing the 12-month measurement period to have the most clinical activity possible. Patient characteristics and criteria for the DMD, ambulation and ventilation algorithms, were captured during a 12-month measurement period surrounding the DBMD diagnosis date (Figure 1). The notes reviewed from the EHR were not limited to this 12-month measurement period and could have been as far back as 2003. The mean number of months with EHR data for the 303 patients who were included in the study was 33.8 months. Patients were excluded if they had missing or invalid age or sex in the claims database. Among patients meeting the inclusion and exclusion criteria, those who were male and aged 40 years or younger on the DBMD diagnosis date were retained in the study sample.
FIGURE 1.
Study Design
CLAIMS-BASED ALGORITHMS
The primary analysis included evaluation of 3 DMD diagnosis algorithms (broad definition, narrow definition, and more restrictive narrow definition), 1 nonambulatory algorithm, and 1 ventilation-dependent algorithm. These 5 algorithms are described in this article. A complete list of codes used for patient identification is given in Supplementary Table 1, available in online article, and a summary of all algorithms evaluated during the algorithm development process is presented as Supplementary Table 2.
The broad DMD definition includes male patients aged 40 years or younger with at least 2 claims with a DBMD diagnosis code (ICD-10-CM G71.01). The narrow DMD definition includes patients meeting the broad DMD definition who also had a prescription for glucocorticoids (betamethasone, budesonide, cortisone, deflazacort, dexamethasone, hydrocortisone, methylprednisolone, prednisolone, prednisone, or triamcinolone) or exon-skipping therapy (casimersen, eteplirsen, golodirsen, or viltolarsen) or evidence of ambulation assistance/support or nonambulatory status at 12 years of age or younger or evidence of ventilation support or dependence; if aged 30 years or older, patients were required to have evidence of ventilation support or dependence. The more restrictive narrow DMD definition includes patients meeting the narrow DMD definition, with the additional requirement for evidence of ventilation support or dependence for those aged at least 20-30 years. The nonambulatory definition includes patients with at least 1 claim for a power wheelchair or bed confinement or at least 1 claim indicating ventilation dependence. The ventilation-dependent definition: includes patients with at least 1 claim for ventilation dependence, chronic or unknown respiratory failure, mechanical ventilation or ventilator, permanent tracheostomy, ongoing care for tracheostomy, or complications of ventilation.
CASE CONFIRMATION
Deidentified provider notes from the EHR data were selected and abstracted to confirm DMD diagnosis and ambulation and ventilation status for patients identified by the claims-based algorithms. For each patient, up to 6 notes during the study period were selected based on note length, recency, and the presence of key terms (eg, DMD, exon skipper, immobility) likely to provide sufficient clinical details of interest (See Supplementary Table 2 for a list of key terms). Patients can have more than 1 note in a single encounter if they saw multiple providers (eg, physician, physical therapist, respiratory specialist), and the number of notes that were available to select depended on the length of follow-up in the EHR (study period), which differs for each patient. The number of notes containing a key term from the patients included in the study ranged from 1 note to 1,324 notes, with a mean of 74 notes and a median of 25 notes. A maximum of 6 notes was selected because it would not have been possible to review all notes, so notes with the most valuable and complete information were prioritized. Tagged concepts (ie, words, grouping of words, sentences) were extracted from the selected notes into a structured data table by clinical experts following a prespecified abstraction guide. The concepts were used to classify patients according to diagnosis (DMD, BMD, DBMD, other or nonspecific MD, non-MD diagnosis, or insufficient notes to determine diagnosis); ambulation status (ambulatory or nonambulatory); and ventilation status (ventilation dependent or nonventilation dependent). A second reviewer then performed quality review of the extracted elements for clarity, accuracy, and consistency. Any differences were adjudicated by a third reviewer.
PATIENT CHARACTERISTICS
Patient characteristics captured from administrative claims data during the 12-month measurement period included age, sex, insurance type, and geographic region; Charlson Comorbidity Index score28 calculated on the basis of diagnosis codes on medical claims; pediatric complex chronic conditions (CCC) index score29 and contributing categories (cardiovascular, gastrointestinal, hematologic or immunologic, malignancy, metabolic, other congenital or genetic defect, renal and urologic, respiratory, technology dependence, transplant); selected mental health and cognitive conditions (ie, ADHD, autism, depression, and anxiety, including OCD and learning disabilities [eg, dyslexia, dyscalculia]); use of DMD-related medications (glucocorticoids, exon-skipping therapy); and indicators of ambulation and ventilation status (Supplementary Table 1). Patient race and ethnicity and whether patients were seen by a primary care provider, cardiologist, or neurologist were captured from the EHR data.
STATISTICAL ANALYSIS
Analyses were conducted using SAS software version 9.4 (SAS Institute, Inc.). Patient characteristics were analyzed descriptively among the subset of patients with provider notes available, stratified by confirmed diagnosis. Performance of the claims-based DMD algorithms was assessed by PPV, calculated as the total number of patients with a confirmed DMD diagnosis in EHR notes (true positives) divided by the total number of patients classified as having DMD by the algorithm (true positives + false positives), with exact binomial 95% CIs provided. PPVs and 95% CIs were generated similarly for the ambulation and ventilation status algorithms.
Some patients whose EHR were abstracted were found to have notes that were not informative enough to determine whether the condition of interest was present (referred to as “insufficient notes”); for example, a patient with only 1 tagged concept stating “breath sounds normal” was categorized as having insufficient notes. PPVs were calculated excluding patients with insufficient notes and were also calculated including patients with insufficient notes to provide a more conservative PPV estimate. The conservative approach assumed that patients with insufficient notes did not have the condition of interest and were classified as non-cases.
Results
STUDY SAMPLE
Among 2,801 patients in the Optum Market Clarity Database with a diagnosis code for DBMD, a total of 1,300 met the additional claims-based inclusion/exclusion criteria. In total, EHR notes were available and abstracted for a sample of 303 patients (Figure 2).
FIGURE 2.
Patient Identification and Attrition
The initial study sample of 1,300 patients had a mean (SD) age of 15.9 (8.7) years, with non-Hispanic White being the largest race and ethnicity group (52.7%). The largest proportion of patients were from the Midwest (34.2%), with the majority having Medicaid (45.1%) or commercial insurance (45%). Most patients had Charlson comorbidity index scores of 0 (57.2%) or 1-2 (30.6%), and the top 3 common pediatric CCC, other than neurologic/neuromuscular disease, were cardiovascular conditions (40.7%), respiratory conditions (21.5%), and technology dependence (17.9%).
Characteristics of patients whose EHR were abstracted (mean [SD] age 14.8 [8.3] years) are shown in Table 1. The race and ethnicity distribution was 61.7% non-Hispanic White, 11.9% Hispanic, 4.3% non-Hispanic Asian, 3.0% non-Hispanic Black, and 19.1% Unknown. The largest proportion of patients was from the Midwest (38.9%), followed by Northeast (27.4%), West (24.8%), South (7.6%), and Other (1.3%). Most patients had Medicaid (50.2%) or commercial insurance (42.9%). The majority of patients had Charlson comorbidity index scores of 0 or 1-2 (59.4% and 27.7%, respectively). Common pediatric CCC other than neurologic/neuromuscular disease included cardiovascular conditions (41.6%), respiratory conditions (24.1%), and technology dependence (20.1%). Mental health and cognitive conditions observed among the study sample included depression, anxiety, or OCD (18.8%); learning disabilities (17.8%); ADHD (10.9%); and autism (8.9%).
TABLE 1.
Characteristics of Patients Whose Electronic Health Records Were Available and Abstracted
| Characteristic | Total (n = 303) | Insufficient notesa (n = 30) | DMD (n = 216) | BMD (n = 34) | DBMD (n = 7) | Otherb (n = 16) |
|---|---|---|---|---|---|---|
| Age, years, mean (SD) | 14.8 (8.3) | 13.1 (9.3) | 14.4 (7.8) | 15.8 (9.3) | 20.0 (8.0) | 19.4 (10.0) |
| Race and ethnicity, n (%) | ||||||
| Non-Hispanic White | 187 (61.7) | 23 (76.7) | 127 (58.8) | 23 (67.7) | 5 (71.4) | 9 (56.3) |
| Hispanic | 36 (11.9) | 2 (6.7) | 27 (12.5) | 2 (5.9) | 1 (14.3) | 4 (25.0) |
| Non-Hispanic Asian | 13 (4.3) | 0 (0.0) | 11 (5.1) | 2 (5.9) | 0 (0.0) | 0 (0.0) |
| Non-Hispanic Black | 9 (3.0) | 2 (6.7) | 7 (3.2) | 0 (0.0) | 0 (0.0) | 0 (0.0) |
| Other/unknown | 58 (19.1) | 3 (10.0) | 44 (20.4) | 7 (20.6) | 1 (14.3) | 3 (18.8) |
| Geographic region, n (%) | ||||||
| Midwest | 118 (38.9) | 14 (46.7) | 77 (35.7) | 18 (52.9) | 5 (71.4) | 4 (25.0) |
| Northeast | 83 (27.4) | 7 (23.3) | 55 (25.5) | 11 (32.4) | 1 (14.3) | 9 (56.3) |
| West | 75 (24.8) | 2 (6.7) | 65 (30.1) | 5 (14.7) | 1 (14.3) | 2 (12.5) |
| South | 23 (7.6) | 5 (16.7) | 17 (7.9) | 0 (0.0) | 0 (0.0) | 1 (6.3) |
| Other | 4 (1.3) | 2 (6.7) | 2 (0.9) | 0 (0.0) | 0 (0.0) | 0 (0.0) |
| Insurance type, n (%) | ||||||
| Medicaid | 152 (50.2) | 11 (36.7) | 114 (52.8) | 16 (47.1) | 4 (57.1) | 7 (43.8) |
| Commercial | 130 (42.9) | 16 (53.3) | 89 (41.2) | 17 (50) | 2 (28.6) | 6 (37.5) |
| Medicare | 6 (2.0) | 0 (0.0) | 4 (1.9) | 0 (0).0 | 1 (14.3) | 1 (6.3) |
| Unknown | 15 (5.0) | 3 (10.0) | 9 (4.2) | 1 (2.9) | 0 (0.0) | 2 (12.5) |
| Charlson Comorbidity Index score category, n (%) | ||||||
| 0 | 180 (59.4) | 22 (73.3) | 122 (56.5) | 20 (58.8) | 6 (85.7) | 10 (62.5) |
| 1-2 | 84 (27.7) | 5 (16.7) | 64 (29.6) | 12 (35.3) | 0 (0.0) | 3 (18.8) |
| 3-4 | 32 (10.6) | 2 (6.7) | 25 (11.6) | 2 (5.9) | 1 (14.3) | 2 (12.5) |
| 5+ | 7 (2.3) | 1 (3.3) | 5 (2.3) | 0 (0.0) | 0 (0.0) | 1 (6.3) |
| Pediatric CCC index score category, n (%) | ||||||
| 0 | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) |
| 1-2 | 202 (66.7) | 24 (80.0) | 139 (64.4) | 21 (61.8) | 6 (85.7) | 12 (75.0) |
| 3+ | 101 (33.3) | 6 (20.0) | 77 (35.7) | 13 (38.2) | 1 (14.3) | 4 (25.0) |
| Pediatric CCC, n (%) | ||||||
| Cardiovascular | 126 (41.6) | 7 (23.3) | 96 (44.4) | 16 (47.1) | 3 (42.9) | 4 (25.0) |
| Respiratory | 73 (24.1) | 3 (10.0) | 62 (28.7) | 4 (11.8) | 0 (0.0) | 4 (25.0) |
| Technology dependence | 61 (20.1) | 4 (13.3) | 51 (23.6) | 3 (8.8) | 0 (0.0) | 3 (18.8) |
| Other congenital or genetic defect | 51 (16.8) | 5 (16.7) | 40 (18.5) | 2 (5.9) | 0 (0.0) | 4 (25.0) |
| Gastrointestinal | 38 (12.5) | 3 (10.0) | 32 (14.8) | 1 (2.9) | 0 (0.0) | 2 (12.5) |
| Metabolic | 38 (12.5) | 2 (6.7) | 21 (9.7) | 12 (35.3) | 1 (14.3) | 2 (12.5) |
| Renal and urologic | 14 (4.6) | 1 (3.3) | 11 (5.1) | 1 (2.9) | 0 (0.0) | 1 (6.3) |
| Hematologic or immunologic | 8 (2.6) | 1 (3.3) | 3 (1.4) | 3 (8.8) | 0 (0.0) | 1 (6.3) |
| Malignancy | 3 (1.0) | 0 (0.0) | 2 (0.9) | 0 (0.0) | 0 (0.0) | 1 (6.3) |
| Transplant | 3 (1.0) | 1 (3.3) | 0 (0.0) | 2 (5.9) | 0 (0.0) | 0 (0.0) |
| Mental health and cognitive conditions, n (%) | ||||||
| Depression/anxiety, including OCD | 57 (18.8) | 6 (20.0) | 42 (19.4) | 4 (11.8) | 2 (28.6) | 3 (18.8) |
| Learning disabilities, including dyslexia/dyscalculia | 54 (17.8) | 8 (26.7) | 34 (15.7) | 7 (20.6) | 1 (14.3) | 4 (25.0) |
| ADHD | 33 (10.9) | 7 (23.3) | 23 (10.7) | 3 (8.8) | 0 (0.0) | 0 (0.0) |
| Autism | 27 (8.9) | 4 (13.3) | 17 (7.9) | 4 (11.8) | 2 (28.6) | 0 (0.0) |
a Electronic health record contained insufficient notes to determine DMD diagnosis status.
b Includes undetermined diagnoses and conditions, such as LAMA2-associated MD, nonspecific limb-girdle MD, limb-girdle/Emery-Dreifuss MD, merosin-negative MD, POMT1 MD, Ulrich MD, and cerebral palsy.
BMD = Becker muscular dystrophy; CCC = complex chronic conditions; DBMD = Duchenne/Becker muscular dystrophy; DMD = Duchenne muscular dystrophy; MD = muscular dystrophy.
Upon review, 55 patients whose EHR notes were available and abstracted were found to have insufficient notes to confirm or rule out the conditions of interest (30 patients had insufficient notes to determine diagnosis and an additional 25 had insufficient notes to determine ambulation/ventilation status).
ALGORITHM PERFORMANCE
DMD Diagnosis. Among patients with sufficient notes available, the more restrictive narrow DMD definition algorithm had the highest PPV; of 143 patients identified with DMD from the claims algorithm, 131 had DMD diagnoses confirmed by EHR notes review, yielding a PPV (95% CI) of 91.6% (85.8%-95.6%) (Table 2). In comparison, the broad DMD definition algorithm identified 202 patients with DMD from the claims data, of which 174 had DMD diagnoses confirmed by EHR notes review (PPV [95% CI] 86.1% [80.6%-90.6%]). The narrow DMD definition algorithm identified 145 patients with DMD, of which 132 had DMD diagnoses confirmed by EHR notes review (PPV [95% CI] 91.0% [85.2%-95.1%]).
TABLE 2.
Algorithm Validation Results: PPV
| Algorithm | Excluding patients with insufficient EHR notesa (n = 273) | Including patients with insufficient EHR notesa (n = 303) | ||||
|---|---|---|---|---|---|---|
| Identified from claims, n | Confirmed in EHR, n | PPV, % (95% CI) | Identified from claims, n | Confirmed in EHR, n | PPV, % (95% CI) | |
| DMD diagnosis | ||||||
| Among all patients | ||||||
| Broad DMD definitionb | 202 | 174 | 86.1(80.6-90.6) | 222 | 174 | 78.4(72.4-83.6) |
| Narrow DMD definitionc | 145 | 132 | 91.0 (85.2-95.1) | 157 | 132 | 84.1(77.4-89.4) |
| More restrictive narrow DMD definitiond | 143 | 131 | 91.6(85.8-95.6) | 155 | 131 | 84.5(77.8-89.8) |
| Excluding patients with insufficient EHR notes (n = 248) | Including patients with insufficient EHR notese (n = 303) | |||||
| Ambulation status | ||||||
| Nonambulatory definitionf | ||||||
| Broad DMD definitionb | 95 | 84 | 88.4(80.2-94.1) | 106 | 84 | 79.3(70.3-86.5) |
| Narrow DMD definitionc | 69 | 59 | 85.5(75.0-92.8) | 79 | 59 | 74.7 (63.6-83.8) |
| More restrictive narrow DMD definitiond | 69 | 59 | 85.5(75.0-92.8) | 79 | 59 | 74.7 (63.6-83.8) |
| Ventilation status | ||||||
| Ventilation-dependent definitiong | ||||||
| Broad DMD definitionb | 45 | 35 | 77.8(62.9-88.8) | 48 | 35 | 72.9(58.2-84.7) |
| Narrow DMD definitionc | 38 | 28 | 73.7 (56.9-86.6) | 41 | 28 | 68.3 (51.9-81.9) |
| More restrictive narrow DMD definitiond | 38 | 28 | 73.7 (56.9-86.6) | 41 | 28 | 68.3 (51.9-81.9) |
b At least 2 claims with a diagnosis code for DBMD, male, and aged 40 years or younger.
c At least 2 claims with a diagnosis code for DBMD, male, and aged 40 years or younger; prescription for glucocorticoids or exon-skipping therapy or evidence of ambulation assistance/support or nonambulatory status at 12 years of age or younger or evidence of ventilation support; and evidence of ventilation support required if aged 30 years or older.
d At least 2 claims with a diagnosis code for DBMD, male, and aged 40 years or younger; prescription for glucocorticoids or exon-skipping therapy or evidence of ambulation assistance/support or nonambulatory status at 12 years of age or younger or evidence of ventilation support; and evidence of ventilation support required if aged 20 years or older.
e For 55 patients, EHR contained insufficient notes; 30 patients had insufficient notes to confirm the DMD diagnosis and an additional 25 patients had insufficient notes to determine ambulation/ventilation status.
f At least 1 claim for a power wheelchair or bed confinement or at least 1 claim indicating ventilation dependence.
g At least 1 claim for ventilation dependence, chronic or unknown respiratory failure, mechanical ventilation or ventilator, permanent tracheostomy, ongoing care for tracheostomy, or complications of ventilation.
DBMD = Duchenne/Becker muscular dystrophy; DMD = Duchenne muscular dystrophy; EHR = electronic health record; PPV = positive predictive value.
Including patients with insufficient notes, the PPV (95% CI) was 84.5% (77.8%-89.8%) for the more restrictive narrow DMD definition algorithm, 78.4% (72.4%-83.6%) for the broad DMD definition algorithm, and 84.1% (77.4%-89.4%) for the narrow DMD definition algorithm.
Nonambulatory Status. Among patients with sufficient notes available, the PPV (95% CI) of the nonambulatory algorithm was 88.4% (80.2%-94.1%) among patients meeting the broad DMD definition (Table 2). The PPVs among patients meeting the narrow DMD definition and the more restrictive narrow DMD definition were identical at 85.5% (75.0%-92.8%). Including patients with insufficient notes, the PPV (95% CI) was 79.3% (70.3%-86.5%) and 74.7% (63.6%-83.8%) among patients meeting the broad and narrow/more restrictive narrow DMD definitions, respectively.
Ventilation-Dependent Status. Among patients with sufficient notes available, the PPV 95% CI of the ventilation-dependent algorithm was 77.8% (62.9%-88.8%) among patients meeting the broad DMD definition and 73.7% (56.9%-86.6%) among patients meeting the narrow DMD definition and the more restrictive narrow DMD definition (Table 2). Including patients with insufficient notes, the PPV (95% CI) was 72.9% (58.2%-84.7%) and 68.3% (51.9%-81.9%) among patients meeting the broad and narrow/more restrictive narrow DMD definitions, respectively.
Discussion
This work presents an administrative claims–based algorithm that improved accuracy in identifying patients with DMD despite the lack of a specific diagnosis code to distinguish DMD from BMD. The more restrictive narrow DMD diagnosis algorithm validated in this study had a PPV of 92%, or 85% as a highly conservative estimate, constituting a substantial increase from the 74% PPV for the algorithm developed prior to the 2018 introduction of ICD-10-CM code G71.01 for DBMD,24 and highlighting the utility of this new code for identifying patients with DMD using claims data alone.
Although the creation of separate diagnosis codes would make it easier to differentiate DMD and BMD in administrative claims, this development is currently unlikely because of the considerable overlap in genetics and clinical presentation for these conditions.1 Given that DMD and BMD have similar signs and symptoms, it can be difficult to definitively diagnose a patient with one or the other; moreover, separating DMD and BMD in medical coding could result in some patients being excluded from approved therapies. This DMD diagnosis algorithm therefore has the potential to become a critical tool for conducting DMD-specific real-world research, as the ability to accurately identify patients with DMD in claims data makes it possible to conduct observational studies to assess disease burden, treatment patterns, and outcomes among patients with DMD treated in clinical practice. This algorithm could also potentially be used in sequence with other DMD-related algorithms—for example, a recently validated algorithm for using claims data supplemented by EHR to identify stages of DMD21—to provide even more nuanced insight into outcomes among this patient population. Considering that patient recruitment for DMD clinical trials is often hampered by the rarity of the condition, the possibility of using this DMD diagnosis algorithm to help identify clinical sites that may have higher numbers of patients with DMD could also be explored.
We developed and evaluated algorithms based on broad, narrow, and more restrictive narrow definitions of DMD. The narrow DMD definition algorithm, which added criteria involving use of DMD-specific therapies, ambulation assistance, and ventilation support, including a requirement for evidence of ventilation support for patients aged 30-40 years, yielded a higher PPV than the broad DMD definition (91% vs 86%). The more restrictive narrow definition, which lowered the age threshold for evidence of ventilation support from 30 to 40 years to at least 20 years, resulted in a slightly improved PPV: 92% vs 91% among patients with sufficient notes available, and 85% vs 84% including patients with insufficient notes. These findings suggest that use of the younger age threshold may be beneficial in future research. However, the best choice of DMD algorithm may vary depending on the research question and the target study population. For example, a study to estimate the burden of disease for patients with DMD would benefit from an algorithm with a higher PPV, with the understanding that some patients with DMD may be excluded. In contrast, a program to identify patients with possible DMD for further assessment and follow-up may be best served by an algorithm that allows a higher amount of misidentification as a trade-off for capturing a larger number of potential patients. The relative complexity of the different algorithms is another consideration. For some purposes, the ease of implementation afforded by use of a simpler algorithm with fewer criteria may be worth some loss of specificity. To aid researchers in choosing the most appropriate algorithm for their particular research goal, a summary of all algorithms developed and evaluated in the course of this study is included as Supplementary Table 2.
In addition to DMD diagnosis algorithms, this work evaluated the performance of algorithms for determining nonambulatory status and ventilation dependence among patients with DMD. Although these algorithms accurately identified the majority of patients with the condition of interest, their PPVs were lower than that of the DMD diagnosis algorithm. The lower PPVs observed for the ambulation and ventilation algorithms (86% and 74%, respectively) are likely in part a function of patient numbers, because PPV is directly related to the prevalence of the target condition in the sample.30 The number of patients who were nonambulatory or ventilation dependent was small relative to number of patients meeting the DMD definition, resulting in a lower PPV.
For all 3 algorithm types, we found that a lack of detailed information about diagnosis and patient ambulation/ventilation status in the EHR presented a challenge during validation. Of the 303 patients identified by the DMD diagnosis claims algorithm, approximately 10% had inadequate notes in the EHR to confirm or rule out DMD diagnosis, and approximately 18% had inadequate notes to determine ambulation/ventilation status. These findings suggest that providers should consider confirming patient diagnosis and describing/documenting disease progression status more consistently at each visit to allow for better management and tracking of patients over time; however, additional documentation would require more of the providers time, which is not always possible, reminding us that claims and EHR data are collected to support health care operations and patient management, and there are limitations to their use for research.
LIMITATIONS
This work has several limitations. First, as with all claims and EHR analyses, some data may be incomplete or contain errors. The degree to which data are incomplete or missing is highly dependent on clinician-level factors, such as workflow, patient volume, and familiarity with the particular EHR system. Second, the sample size was relatively small (n = 303) because not all patients identified in the claims (n = 1,300) had EHR notes available for algorithm validation, and genetic test results and other clinical/biometric measures useful for confirming a DMD diagnosis or identifying ventilation/ambulation status were available for only a subset of patients (n = 273 and n = 248, respectively). Younger and/or newly diagnosed patients are more likely to have been missed, as they may not yet have begun some of the treatments and equipment use that were incorporated as algorithm criteria. However, we found that PPV among younger patients in this study was generally higher than or similar to that among older patients (Supplementary Table 2). Third, patients who were not receiving regular care from providers included in the EHR database or who were diagnosed by providers not included in the EHR database may have been misclassified as not having DMD because of lack of information; however, as only 8 of 273 patients (2.9%) had an indeterminate MD diagnosis, the impact of this limitation was likely minimal. Finally, results may not be generalizable to uninsured populations because the study population was selected from among patients with commercial, Medicare Advantage, or Medicaid insurance.
Conclusions
This work provides the means to more accurately identify patients with DMD when using an ICD-10-CM diagnosis code that is nonspecific to DMD. The algorithms validated in this study can be applied to assess treatment effectiveness and other outcomes among real-world patients with DMD in administrative claims databases.
ACKNOWLEDGMENTS
The authors thank Sarangi Bhalani, MBBS, MBA, and Saniya Choudhary, MBBS (both of Optum), for EHR notes abstraction; Felix Cao, PhD, and Damon Van Voorhis, BS (both of Optum), for programming; Michael Johnson, MS (Optum) for analysis consultation; Yvette Edmonds, PhD (Optum), and Deja Scott-Shemon, MPH (Optum), for medical writing assistance; Jessica Fachini, MBA (Optum), for project management; and Sumit Jhamb, BPharm (Optum), for project support.
REFERENCES
- 1.Mercuri E, Bonnemann CG, Muntoni F. Muscular dystrophies. Lancet. 2019;394(10213):2025-38. doi:10.1016/S0140-6736(19)32910-1 [DOI] [PubMed] [Google Scholar]
- 2.Romitti PA, Zhu Y, Puzhankara S, et al. Prevalence of Duchenne and Becker muscular dystrophies in the United States. Pediatrics. 2015;135(3):513-21. doi:10.1542/peds.2014-2044 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Mah JK, Korngut L, Dykeman J, Day L, Pringsheim T, Jette N. A. systematic review and meta-analysis on the epidemiology of Duchenne and Becker muscular dystrophy. Neuromuscul Disord. 2014;24(6):482-91. doi:10.1016/j.nmd.2014.03.008 [DOI] [PubMed] [Google Scholar]
- 4.Sheehan DW, Birnkrant DJ, Benditt JO, et al. Respiratory management of the patient with Duchenne muscular dystrophy. Pediatrics. 2018;142(Suppl 2):S62-71. doi:10.1542/peds.2018-0333H [DOI] [PubMed] [Google Scholar]
- 5.Ciafaloni E, Kumar A, Liu K, et al. Age at onset of first signs or symptoms predicts age at loss of ambulation in Duchenne and Becker muscular dystrophy: Data from the MD STARnet. J Pediatr Rehabil Med. 2016;9(1):5-11. doi:10.3233/PRM-160361 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Lo Mauro A, Aliverti A. Physiology of respiratory disturbances in muscular dystrophies. Breathe (Sheff). 2016;12(4):318-27. doi:10.1183/20734735.012716 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Szabo SM, Salhany RM, Deighton A, Harwood M, Mah J, Gooch KL. The clinical course of Duchenne muscular dystrophy in the corticosteroid treatment era: A systematic literature review. Orphanet J Rare Dis. 2021;16(1):237. doi:10.1186/s13023-021-01862-w [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Andrews JG, Wahl RA. Duchenne and Becker muscular dystrophy in adolescents: Current perspectives. Adolesc Health Med Ther. 2018;9:53-63. doi:10.2147/AHMT.S125739 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Verhaart IE, Aartsma-Rus A. Therapeutic developments for Duchenne muscular dystrophy. Nat Rev Neurol. 2019;15(7):373-86. doi:10.1038/s41582-019-0203-3 [DOI] [PubMed] [Google Scholar]
- 10.Waldrop MA, Flanigan KM. Update in Duchenne and Becker muscular dystrophy. Curr Opin Neurol. 2019;32(5):722-7. doi:10.1097/WCO.0000000000000739 [DOI] [PubMed] [Google Scholar]
- 11.Ryder S, Leadley RM, Armstrong N, et al. The burden, epidemiology, costs and treatment for Duchenne muscular dystrophy: An evidence review. Orphanet J Rare Dis. 2017;12(1):79. doi:10.1186/s13023-017-0631-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Birnkrant DJ, Bushby K, Bann CM, et al. Diagnosis and management of Duchenne muscular dystrophy, part 2: Respiratory, cardiac, bone health, and orthopaedic management. Lancet Neurol. 2018;17(4):347-61. doi:10.1016/S1474-4422(18)30025-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Birnkrant DJ, Bushby K, Bann CM, et al. Diagnosis and management of Duchenne muscular dystrophy, part 1: Diagnosis, and neuromuscular, rehabilitation, endocrine, and gastrointestinal and nutritional management. Lancet Neurol. 2018;17(3):251-67. doi:10.1016/S1474-4422(18)30024-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Matthews E, Brassington R, Kuntzer T, Jichi F, Manzur AY. Corticosteroids for the treatment of Duchenne muscular dystrophy. Cochrane Database Syst Rev. 2016;(5):CD003725. doi:10.1002/14651858.CD003725.pub4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.McDonald CM, Henricson EK, Abresch RT, et al. Long-term effects of glucocorticoids on function, quality of life, and survival in patients with Duchenne muscular dystrophy: A prospective cohort study. Lancet. 2018;391(10119):451-61. doi:10.1016/S0140-6736(17)32160-8 [DOI] [PubMed] [Google Scholar]
- 16.Wilton-Clark H, Yokota T. Recent trends in antisense therapies for Duchenne muscular dystrophy. Pharmaceutics. 2023;15(3):778. doi:10.3390/pharmaceutics15030778 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Parent Project Muscular Dystrophy. Duchenne drug development pipeline. Accessed June 29, 2022. https://www.parentprojectmd.org/duchenne-drug-development-pipeline/ [Google Scholar]
- 18.Schulz KF, Altman DG, Moher D, Group C.. CONSORT 2010 statement: Updated guidelines for reporting parallel group randomised trials. Int J Surg. 2011;9(8):672-7. doi:10.1016/j.ijsu.2011.09.004 [DOI] [PubMed] [Google Scholar]
- 19.Nazha B, Yang JC, Owonikoko TK. Benefits and limitations of real-world evidence: Lessons from EGFR mutation-positive non-small-cell lung cancer. Future Oncol. 2021;17(8):965-77. doi:10.2217/fon-2020-0951 [DOI] [PubMed] [Google Scholar]
- 20.Smith MG, Royer J, Mann JR, McDermott S. Using administrative data to ascertain true cases of muscular dystrophy: Rare disease surveillance. JMIR Public Health Surveill. 2017;3(1):e2. doi:10.2196/publichealth.6720 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Iff J, Zhong Y, Gupta D, et al. Disease progression stages and burden in patients with Duchenne muscular dystrophy using administrative claims supplemented by electronic medical records. Adv Ther. 2022;39(6):2906-19. doi:10.1007/s12325-022-02117-1 [DOI] [PubMed] [Google Scholar]
- 22.Thayer S, Bell C, McDonald CM. The direct cost of managing a rare disease: Assessing medical and pharmacy costs associated with Duchenne muscular dystrophy in the United States. J Manag Care Spec Pharm. 2017;23(6):633-41. doi:10.18553/jmcp.2017.23.6.633 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Soslow JH, Hall M, Burnette WB, et al. Creation of a novel algorithm to identify patients with Becker and Duchenne muscular dystrophy within an administrative database and application of the algorithm to assess cardiovascular morbidity. Cardiol Young. 2019;29(3):290-6. doi:10.1017/S1047951118002226 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Chen X, Agiro A, Martin AS, Lucas AM, Haynes K. Chart validation of an algorithm for identifying hereditary progressive muscular dystrophy in healthcare claims. BMC Med Res Methodol. 2019;19(1):174. doi:10.1186/s12874-019-0816-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Mallow P, Belk K, Clark L, Banuelos R, Martin C. POSC244 racial disparities in healthcare utilization outcomes among patients with Duchenne muscular dystrophy. Value Health. 2022;25(1):S167. doi:10.1016/j.jval.2021.11.811 [Google Scholar]
- 26.Ayme S, Bellet B, Rath A. Rare diseases in ICD11: Making rare diseases visible in health information systems through appropriate coding. Orphanet J Rare Dis. 2015;10:35. doi:10.1186/s13023-015-0251-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Kennedy A. New Duchenne/Becker ICD code in effect today—G71.01. Accessed September 17, 2022. https://www.parentprojectmd.org/new-duchenne-becker-icd-code-in-effect-today-g71-01/
- 28.Quan H, Li B, Couris CM, et al. Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries. Am J Epidemiol. 2011;173(6):676-82. doi:10.1093/aje/kwq433 [DOI] [PubMed] [Google Scholar]
- 29.Feudtner C, Feinstein JA, Zhong W, Hall M, Dai D. Pediatric complex chronic conditions classification system version 2: Updated for ICD-10 and complex medical technology dependence and transplantation. BMC Pediatr. 2014;14:199. doi:10.1186/1471-2431-14-199 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Monaghan TF, Rahman SN, Agudelo CW, et al. Foundational statistical principles in medical research: Sensitivity, specificity, positive predictive value, and negative predictive value. Medicina (Kaunas). 2021;57(5):503. doi:10.3390/medicina57050503 [DOI] [PMC free article] [PubMed] [Google Scholar]