Free text

Other important considerations in the design of an IRS are the data storage and analysis requirements. These need to be considered early in the design of the system if it is to be used to research and display trends effectively. For example, in addition to the answers to specific questions, the accident data analyst may wish to make use of free text descriptions of the circumstances of the accident. This implies that a text-based retrieval system will be required. 

Scientific information—the contextual interpretation of experimental data—is published as free text. The same applies to the annotation of experimental results, genes, proteins, and compounds and the description of medical conditions. This clearly indicates that scientific information is not structured, which creates a major challenge for its reuse, management, and statistical analysis. This fact has largely been recognized, and much research... 

As set forth above, terms are identified from free text and database records by text mining. The terms are then associated with relevant concepts, which... 

The problem for the statistical programmer in categorizing data comes from text variables, or more specifically, free-text variables. A free-text variable is one that may contain any characters and is typically limited only in length. As an example, let s say you need to summarize the adverse events for a set of patients in a trial. The following SAS code shows the data and a quick summarization of the adverse events. 

Program 2.2 Summarizing Free-Text Adverse Event Data... 

There are three problems with this adverse events summary. First, HEADACHE and HEDACHE are counted as separate events even though it is clear that the latter is simply a misspelling of the former. Second, MI and MYOCARDIAL INFARCTION are considered as separate events even though the former is simply an abbreviation of the latter. Finally, LIGHTHEADEDNESS/FACIAL LACERATION refers to perhaps related but different adverse events that need to be counted separately. All three of these problems exist because the data were entered in free-text fashion. 

There is only one good solution to handling free-text variables that are needed for statistical analysis. The free-text variables need to be coded by clinical data management in the clinical database. If the adverse events were coded with a dictionary such as MedDRA, the previous example might look like Program 2.3. 

In summary, for data to be useful in clinical trial analyses they need to be quantifiable. The data must be either a continuous measure or a categorical value. Free text poses a problem for analysis, and if it is a valuable variable for the statistical analyses it really must be coded. Finally, hardcoding should be used only when absolutely necessary, because it is inherently problematic. Organizations that do allow hardcoding should document in their standard operating procedures (SOPs) that it is an approved business practice and how it is to be used. 

Concomitant medications and prior medications are collected in one of two forms a list-type free-text format where the medications get coded later by data management, or a pre-categorized data format. Here is the free-text CRF format ... 

Again, the free-text CRF format is useful in that it allows for explicit description of the historical condition, whereas the pre-categorized CRF format omits that detail. However, the free-text list format necessitates coding with a coding dictionary such as MedDRA in order to be useful for analyses. The pre-categorized format is useful here, as only medical history relevant to the investigational therapy can be captured and the cost of additional coding of the history data is eliminated entirely. 

The adverse event form is fairly standard across clinical trials. The form consists of a list of events for which data are entered as free text and are later coded with a dictionary such as MedDRA and some associated event attribute variables. In just about any clinical trial, an adverse event form very similar to the following sample will be found. 

Data management in RS/1 is based on two-dimensional tables. Each cell of a table can contain data representing fixed or floating point numbers, dates, times, or free text. Cells in a particular column are not all constrained to the same type it is possible, for example, to include a note about some missing data in a column of numerical results. A user can work with many hundreds of tables. Tables are based on disk files, accessed through a kind of paging scheme, so there is no limit on table size. Some users work with tables containing hundreds of columns and tens of thousands of rows. 

Printed on acid-free text paper, manufactured in accordance with ANSI/ NISO Z39.48-1992 (Permanence of Paper). 

As mentioned earlier, chemical composition and identification information should already have been obtained before the chemical is to be searched. With most information retrieval systems this is a relatively straightforward procedure. Citations on a given subject may be retrieved by entering the desired free text terms as they... 

These data do not change very often, should not be entered as free text, and should introduce very little overhead into the overall registration process. It is recommended that they be cached in the application and refreshed periodically. This way, they can be accessed quickly from the memory rather than queried every time from the underline database. In memory data access is much faster than any input/output (I/O) operations, especially I/O that involves network traffic. As the dictionary data do not change very often, the performance benefit of accessing it from the cache overweighs the benefit of real-time up-to-date data. 

On TOXLINE, field names are identified by a two-letter mnemonic. A list of ETIC fields and their field identifiers on TOXLINE appears in Table 10. If the two-letter identifier is not used in searching, retrieval defaults to free-text searching of an index created by single words from all searchable fields. Thus, if one does not use identifiers, a search for acetylsalicylic acid must be worded ETIC AND ACETYLSALICYLIC AND ACID. Individual words in only the TI, KW, and AB fields are searchable via text words (TW). Indexing terms in most of the technical fields have been put into the keyword (KW) field. 

Open questions These allow free text response, e.g. 

L. The nature of the information collected is it just descriptive reports or also causal Is it just free text, or does it include also answers to a standard set of questions ... 

Substituents marked with an asterisk should not be used in constructing formal names, either because they are definitely obsolete or because they are informal descriptors often used in free text but not approved for constructing actual names (e.g., aryl, brosyl). Apart from these, the list does not give a definite preference for one alternative over another, except in a few cases (e.g., caproyl), which should definitely be avoided because of inaccuracy or ambiguity. Different publications, including CAS, have different editorial preferences. 

A search of two major data bases (MEDLINE, International Pharmaceutical Abstracts) was conducted to identify articles published between January 1988 and December 1995. The beginning date of January 1988 was selected because the original ACCP prospectus was inclusive through December 1987. Both MeSH and free text search terms were used to identify English language articles assessing the value of clinical pharmacy services. Search terms were clinical pharmacy services, pharmacy... 

---