C O D I N G M A N A G E M E N T Introduction Along with the Clinical Trial data, CDM collects data to be coded Coding and Dictionary maintenance handled by CDM group Specialized coding group- Thesaurus Management Group Organization’s Drug Safety Group also involved in Adverse Event related coding “ Medical Monitors, Clinical Scientists, Statisticians analyze medication and adverse event data for safety and efficacy study “ “ “ “ Coding “ The process of grouping or classifying reported terms like medications, adverse events, medical history, diagnoses, disease conditions with reference to known standard terms Coding? grouping or classifying Dictionaries ” What Are They? Diagnoses Diseases Medications Medical History “ An English Dictionary has a list of words “ List of possibilities or standard terms is a dictionary or thesaurus “ CDM Dictionaries have listings of Diagnoses/Diseases, Drugs etc., either pre-listed or can incorporate newer terms (Autocoders) “ Coding Dictionaries can be updated regularly to include newer terms, which is a tedious process Coding Dictionaries “ Dictionaries used in processing, analyzing, and reporting of data collected in clinical trials “ They vary from simple code lists to large complex dictionaries “ Some dictionaries used for years, relatively stable “ Recent dictionaries can be updated regularly Overview of Coding Process AE or Medication entered in Database (Verbatim text) Batch run of TMS Verbatim text matched with Dictionary Pattern Matching/link to similar dictionary term Assign new Preferred Term update dictionary after approval Exact Match found ? Yes No Code assigned Flag for CDM Team Referred to Specialist If Coding Fails…. “ Discrepancies arise if the terms do not get coded “ These discrepancies should be managed i.e., appropriate action taken by Data Management “ Feedback can be given by the Thesaurus Management Group, about appropriate action to be taken “ Based on proper feedback, edits or updates of the stored data should be made in the database Structure of Dictionaries “ Different dictionaries follow different structures based on the classification levels and depth of the classification “ A simple structure for a dictionary usually includes a ‘Preferred Term’ (for a matched verbatim term) and a ‘Code’ “ Text term is always known as the Preferred Term “ A code is ‘Numeric in nature’, but not a number always “ A code can be a text, usually a version of the preferred term Different Forms of Coding “ Electronic versus Paper version of coding dictionaries Electronic Terms are arranged in tables consisting of drugs, events, and diagnoses System refers to these tables & assigns codes Example: Oracle TMS- Thesaurus Management System Paper Like Terms are grouped together Coding specialist looks for a suitable match Code is assigned Understanding the Need for Coding “ Consider a Clinical Trial for Safety and efficacy evaluation for Headache for a investigational pain reliever “ Statisticians would be interested in analyzing the frequency of taking the OTC pain reliever before, during, or after the investigational drug was taken “ Also, number of attacks of headache (adverse event) is analyzed “ The patient might take several OTC drugs which may have the same active ingredient/ingrediens “ Hence, if these medications are not ‚coded‛ for standard terms, this data will significantly affect the Safety & Efficacy Data More Examples ParacetamolGeneric Name “ “ “ “ versus CrocinTrade Name Example: Paracetamol and Crocin Generic name versus Trade name of a medication Both these medications have the same active ingredient ‘Paracetamol’ If different codes are assigned to each of these medications, during review/analysis of drug, safety and efficacy will be affected Need for Coding terms in CDM “ Consider an example for adverse event “ Example: ‘Vomiting’ versus ‘Mild Vomiting’ “ Both events resemble, but the verbatim differs and hence do not have an exact match with the dictionary term “ Differences in analysis will arise if same code is not assigned to each of these adverse events. Versus Vomiting Mild Vomiting Minimum Standards Select appropriate dictionaries that meet project requirements Install and securely maintain dictionaries Implement an audit trail for all changes to the coding dictionary Identify dictionary and version in clinical study reports and integrated summaries “ Store all versions of dictionaries for future reference “ “ “ “ Few Best Practices “ Select auto-coder to facilitate consistent use of dictionaries “ Include version of dictionary “ Store data from all coded dictionary levels whether or not all are initially analyzed or reported “ Establish process for evaluating a change in a version change “ Use of same version of dictionary for coding in combined studies “ Educate other individuals involved in the analyzing and reporting coded data on the functionality and capabilities of dictionaries used Process of Coding Collection of terms No Yes Store the terms Manual review and update the term Matching of terms Match found How Organizations Handle Coding? “ Small organizations “ Big organizations Coding handled by CDM group Separate Coding Specialists or Thesaurus Management Groups Challenges While Recording of Adverse Events/diagnosis/diseases “ Mis-spelling of certain terms “ Numbers and text instead of standard medical terms “ Recording phrases or sentences instead of a standard term “ More than one/multiple adverse events recorded together “ Using certain symbols example: Fever ‚&‛ rashes “ Listing Surgical Procedures as events Record only standard medical terms Dictionary Maintenance/Updates “ Maintenance of dictionaries presents multiple challenges New version dictionary… “ New versions/updates have influence on needs proper care !! already coded terms “ Coding groups should decide well in advance how to handle version changes “ Changes have an impact on adverse events/drugs which already coded Dictionary Types Coding Symbols for a Thesaurus of Adverse Reaction Terms COSTART ICD-9-CM Dictionaries International Classification of Diseases, Ninth Revision, Clinical WHOART World Health Organization Adverse Modification Reactions Terminology Medical Dictionary for Regulatory Activities MedDRA WHODRL World Health Organization Drug Reference List Dictionary Change Control “ Selected dictionary and version should be documented for future reference “ Coding dictionary securely maintained “ If modifications are made, it should be clearly stated - so as not to mislead reviewers “ Changes made to dictionary entries should be included in audit trial “ Modifying published dictionaries discouraged by ICH Dictionary Version Control “ Dictionaries may be available in an electronic or printed format and multiple versions released or published “ These versions for a given project, time period or sets of data clearly documented “ Evaluation of extent of changes between versions “ Impact on previously coded terms “ Criteria for implementing the latest version and recording Selection of a Dictionary-Things to Keep in Mind “ The needs of all users that process, analyze, and report coded data “ Regulatory agencies who eventually review the information Have we thought about the needs of all these members…? Auto-coders/Auto-encoders “ Most modern/advanced coding dictionaries heavily depend on auto-coders “ Broadly may be classified into two types: ” Simple (use exact-match algorithms) auto-coders ” Sophisticated (use multi-step algorithms) auto-coders “ Algorithms are logical step-by-step instructions written for coding Our auto-coders take care … no worries…. Coding Specialist Advantages of Auto-coders “ “ “ “ Re-evaluation of previously coded terms taken care Consistency in coding Capable of handling large numbers of entries Coding Algorithm (set of instructions) ease handling of misspellings, synonyms, word variations etc. Exact-match Algorithms “ Exact-match algorithms work well for medications rather than adverse events “ Variations for medications include mis-spellings, spelling variations “ Few Advanced auto-coders can store the unmatched term, and the code assigned can become more effective “ Over a period of time, the percentage of unmatched terms is reduced “ Since adverse events have more variations during recording, they need more complex algorithms to be effective Multi-step Algorithms “ Unlike exact-match algorithms which help in matching terms with various additional tables, multi-step algorithms perform to distinct operations “ ‘Text transformation’ and ‘Lookups’ “ Success of these algorithms are dependant on the types of transformations and the order of the lookups designed “ Text transformations written must be written in conjunction with the dictionary tables stored in the system Example of a Multi-step Algorithm Step Number 1 2 3 4 5 6 7 Type of Step Capitalize text Remove extra characters and spaces Remove non-contributing words Replace words and phrases Look up in Main dictionary table Look up in dictionary synonym table Look up in company term table Frequently Used Text Transformations “ “ “ “ “ “ “ “ Capitalization Removal of selected punctuation, symbols (example: ‚Cough‛) Removal of extra spaces Removal of numbers (example: adverse event ‚102.1 fever‛) Removal of 'numeric words (example: first, second) Removal of dates and date words Removal of non-contributing words (example: the, patient, complained) Replacement of words with standard terms Lookup Procedures in Multi-step Algorithms “ “ “ “ “ Lookup usually happens after text transformations Alternatively, intermediate lookups can be accomplished If large dictionary tables have to be searched, lookup time is increased Hence, most companies prefer coding being carried out in ‘Batch’ Coding done interactively is less encouraged since it slows down the process Example of Lookup Procedures Step Number Type of Step 1 Capitalize text 2 Remove extra characters and spaces 3 Remove non-contributing words 4 Replace words and phrases 5 Look up in Main dictionary table 6 Look up in dictionary synonym table 7 Look up in company term table Lookup procedure
"Medical coding in Clincial Trial"