Naturalistic Queries
From MindLinc Wiki
Retrospective Data Analysis
Contents |
Clinical Research Information System (CRIS)
The Clinical Research Information System (CRIS) began at Duke University Medical Center in 1998 with the development of an electronic medical record (EMR) to document patient visits in the Department of Psychiatry. The various aspects of a patient's treatment and information about the patient are stored on the EMR. An electronic medical record of the patient's visit provides the health care provider with a readily available means of monitoring the patient's course of treatment. CRIS is an organic system, as new information is added and refinements made to the system regularly.
Nature of the Data
The EMR documents various aspects of a patient's visit and stores the information in distinct files. As each of the aspects of a patient's history—whether it be medicines, diagnosis, demographic characteristics of the patient, treatment outcomes or adverse event, is measured in unique units, data from each facet are stored in separate files which may be merged with each other to provide a complete record of the treatment. Patient data are linked by an ID number and visit number.
The Analytic Sample
In order to comply with HIPAA requirements regarding the analysis of medical records, the data in this database are anonymized. This process permits the analysis of the data without a requirement of securing approval from the Institutional Review Board. In this process, all indications of a patient's identity are either removed (e.g. name) or modified (e.g. dates) to prevent any ability to link the data to the actual person. The data are anonymized periodically, and new generations produced to replace earlier ones. Each generation is a superset of each preceding one, though the anonymization process prevents any linking across generations of data. Not ever entry in CRIS is included in the analytic sample. Only those visits identified through billing codes as involving medication management or therapy are included. Furthermore, only those visits occurring since January 1, 2001, are included. The current analytic sample runs from January, 2001, to April, 2008. Data Sources As medical centers employ the CRIS system, data from visits to those institutions will be added. As of 2/1/2008 the current sample includes over 83,000 patients and 1,100,000 visits
Structure of the Data
The data consist of several files, not unlike the form in which it is originally collected. In the interest of efficiency, where it is possible to combine files into a single one, this was done. The different files have different units of analysis. Recognizing the appropriate unit of analysis for the files is essential to a proper merger of files. The basic unit of analysis in the database is the visit. Where data can be easily represented with the visit as the unit of analysis, this data are combined in a single file. For other data, it is preferable to maintain separate files, each with its own unit of analysis.
The VISITS file.
This file contains a record of each qualified visit in the analytic sample. The unit of analysis is the patient-visit. The VISITS file contains as much information about the patient and his or her treatment as can reasonably be contained on a single record. Besides the ID number of the patient, the ID number and date of the visit, the file contains information regarding the billing codes, type of visit, demographic characteristics of the patient, the primary psychiatric diagnosis, severity of the illness and a measure of treatment efficacy. Five types of visits are recorded: inpatient, outpatient, emergency room, phone notes, and treatment sessions in a drug and alcohol program.
The demographic characteristics of the patients are age at time of the visit, gender and race. Race is classified as white, black, other and not-ascertained (missing).
Psychiatric Diagnoses
The psychiatric diagnosis of the patient is noted at each visit. The diagnoses are coded according to the DSM-IV code. As it is possible for a patient to have more than one diagnosis, each diagnosis is recorded on a unique record. Thus the unit of analysis in this file is the patient-visit-diagnosis. The 412 DSM-IV diagnoses are classified according to 66 subcategories, which are then further classified according to 18 categories. The subcategory is the primary diagnostic category for analyses. The table below presents the diagnostic subcategories, the diagnostic category with which each is associated and the DSM-IV diagnoses associated with each subcategory. The primary diagnosis is noted at each visit. One entry per visit is the appropriate form of this measure. The primary diagnosis of a patient is taken to be the last primary diagnosis entered for the patient during the course of his or her treatment. Both the primary diagnosis noted at the visit and that assigned on the basis of the final diagnosis are included in the psychiatric diagnoses data file.
Medications
There are two medications files: one for psychiatric medications and one for all other medications reported by the patient. Whereas the former is a primary area of concern in the analysis of the database, the latter is of limited utility. Nevertheless should information about non-psychiatric medication use be desired, the two files may be easily joined by concatenating them. The psychiatric medication file contains the following variables: SOURCE--Source of the data (Duke or Texas) PATID--Patient ID Number EDNUM--Visit Number DATE--Date of the visit MEDICATION--Name of the medication GENERIC--Generic name of the medication GENERIC_ID--Generic ID of the medication RXSUBCAT--Subcategory of the medication RXSUBCAT_ID--Subcategory Id RXCAT--Category of the medication RXCAT_ID--Category Id DOSE--Dosage of the medication REMIMEN--How often the medication is taken ROUTE--Route by means the medication is taken DISCONTINUE_REASON--Reason given for discontinuing a drug
In the 2008 generation of the data, 174 psychiatric medications are identified. These are grouped into 33 subcategories of medications, which are in term grouped into 15 categories. The table below presents the subcategory, the category to which it belongs and the generic medicines included in that subcategory. (Table goes here)
The dosage information enables one to tracking the dosing of the medications throughout the course of treatment. A separate dictionary enables one to calculate the total daily dose from the DOSE and REGIMEN. In the 2008 generation of data, the calculation of total daily dose is complete only for the anti-depressant medications. Extending the dictionary to other classes of medications is one aspect of the continuing evolution of the data.
Severity and Outcome Measures: The CGI Scales
The Clinical Global Index (CGI) scales are commonly employed as the measures of the severity of the illness and the efficacy of the treatment. The database contain two CGI measures, severity (CGI-S) and improvement (CGI-I). They are contained in the VISITS file.
The CGI-Severity scale is numeric and has the following values:
- 1=Normal, not at all
- 2=Borderline mentally ill
- 3=Mildly ill
- 4=Moderately ill
- 5=Markedly ill
- 6=Severely ill
- 7=Among the most extremely ill patients
- 8=Missing
In prior analyses, a common practice was to consider those with CGI-S scores of 4 through 7 as “sick”. The CGI-Improvement scale is also numeric and has the following values:
- 0='Not assessed'
- 1='Very much improved'
- 2='Much improved'
- 3='Minimally improved'
- 4='No change'
- 5='Minimally worse'
- 6='Much worse'
- 7='Very much worse'
- 8='Not assessed (first visit)'
- 9='Not assessed'
- 10='Missing (first visit)'
- 11='Missing’
The CGI_I scale is the major measure of treatment efficacy. In prior analyses, scores of 1-2 were said to represent significant improvement in the patient’s condition.
Adverse Events file
A major concern in the course of treatment is the ability of the patient to tolerate his or her medications. Medicine side effects, i.e. adverse events, are recorded at each visit. Since a person may have more than one side effect noted on a visit, the unit of analysis is the patient-visit-side effect. In the 2008 data, 2045 unique side effects are present. The number is as large as it is because physicians use free-text to note the side effect. These 2045 side effects are grouped into 87 subcategories, the major analytic measure of adverse events, which are in turn grouped into 18 categories. A table presenting the subcategories, the category with which each is associated and some representative side effects of each subcategory is presented below.
