Question

Health records are classified as primary or secondary records. Why is this the case? Then, relate this to the clinical data in a patient’s chart. What is the correlation? Give an example of each.

Answer 1

1) With the growing availability of large electronic health record(EHR) databases, clinical researchers are increasingly interested in the secondary use of clinical data. While the prospective collection of data is notoriously expensive and time consuming, the use of an EHR may allow a medical institution to develop a clinical data repository containing extensive records for large numbers of patients, thereby enabling more efficient retrospective research. These data are a promising resource for comparative effectiveness research, outcomes research, epidemiology, drug surveillance, and public health research.
Unfortunately, EHR data are known to suffer from a variety of limitations and quality problems. The presence of incomplete records has been especially well documented. The availability of an electronic record for a given patient does not mean that the record contains sufficient information for a given research task.Data completeness has been explored in some depth. The statistics community has focused extensively on determining in what manner data are missing. Specifically, data may be considered to be missing at random, missing completely at random, or missing not at random. Datasets that meet these descriptions require different methods of imputation and inference.The statistical view of missing or incomplete data, however, is not sufficient for capturing the complexities of EHR data. EHR records are different from research data in their methods of collection, storage, and structure. A clinical record is likely to contain extensive narrative text, redundancies (i.e., the same information is recorded in multiple places within a record), and complex longitudinal information. While traditional research datasets may suffer from some degree of incompleteness, they are unlikely to reflect the broad systematic biases that can be introduced by the clinical care process.There are several dimensions to EHR data completeness. First, the object of interest can be seen as the patient or as the health care process through which the patient was treated; there is a difference between complete information about the patient versus complete information about the patient’s encounters. A patient with no health care encounters and an empty record has a complete record with respect to the health care process, but a blank one with respect to the patient. Furthermore, one can measure completeness at different granularities: the record as a whole or of logical components of the record, each of which may have its own requirements or expectations (e.g., demographic patient information versus the physician thought process). Another dimension of completeness emerges from the distinction between intrinsic and extrinsic data requirements. One can imagine defining minimum information requirements necessary to consider a record complete (which could be with respect to either the patient or the health care process), or one can tailor the measurement of completeness to the intended use. Put another way, we can see completeness in terms of intrinsic expectations (i.e., based a priori upon the content) or extrinsic requirements (based upon the use).The EHR data consumers who define these extrinsic requirements will have different data needs, which will in turn dictate different conceptualizations of a complete patient record. Here, Juran’s definition of quality becomes valuable: 'fitness for use'. It may be that data completeness does not have a simple, objective definition, but is instead task-dependent. Wang and Strong, for example, in their work developing a model of data quality, define completeness as the extent to which data are of sufficient breadth, depth, and scope for the task at hand. In other words, whether a dataset is complete or not depends upon that dataset’s intended use or desired characteristics. In order to determine the number of complete records available for analysis one must first determine what it means to have a complete patient record. The quality of a dataset can only be assessed once the data quality features of interest have been identified and the concept of data quality itself has been defined.Multiple interpretations of EHR completeness, in turn, may result in different subsets of records that are determined to be complete. The relationships between research task, completeness definition, and completeness findings, however, are rarely made explicit. Hogan and Wagner offer one of the most widely used definitions: the proportion of observations that are actually recorded in the system. This definition does not, however, offer specific measures for determining whether a record is complete. Neither does it account for the possibility that completeness may be task dependent. What proportion of observations should be present? Which observations are desired? Are there any other considerations beyond simple proportion? Furthermore, observations are complex, nested concepts, and it must be determined what level of detail or granularity is needed or expected. In order of increasing detail, one could record a visit that occurred, the diagnoses, all the symptoms, a detailed accounting of the timing of all the symptoms, the clinician's thought process in making a diagnosis, etc.In the sections below, we enumerate four specific operational and measurable definitions of completeness. These definitions are not exhaustive, but they illustrate the diversity of possible meanings of EHR data completeness. We ran the definitions against our clinical database in order to demonstrate the magnitude of completeness in the database and to illustrate the degree of overlap among the definitions.

Materials and methods :-

Previously, we conducted a systematic review of the literature on EHR data quality in which we identified five dimensions of data quality that are of interest to clinical researchers engaged in the secondary use of EHR data. Completeness was the most commonly assessed dimension of data quality in the set of articles we reviewed. Based upon this exploration of the literature on EHR data quality, consideration of potential EHR data reuse scenarios, and discussion with stakeholders and domain experts, we describe four prototypical definitions of completeness that represent a conceptual model of EHR completeness. Further definitions of completeness are possible and may become apparent as the reuse of EHR data becomes more common and more use cases and user needs are identified.presents a visual model of the four definitions of completeness, which are described further in Section.
Documentation: a record contains all observations made about a patient.The most basic definition of a complete patient record described in the literature is one where all observations made during a clinical encounter are recorded. This is an objective, task independent view of completeness that is, in essence, a measure of the fidelity of the documentation process. Assessments of documentation completeness rely upon the presence of a reference standard, which may be drawn from contacting the treating physician, observations of the clinical encounter or comparing the EHR data to an alternate trusted data source often a concurrently maintained paper record. Documentation :-completeness is also relevant to the quality measurements employed by the Centers for Medicare & Medicaid Services.In secondary use cases, however, the data consumer may be uninterested in the documentation process. Instead, completeness is determined according to how well the available data match the specific requirements of the task at hand, meaning that completeness in these situations is more often subjective and task dependent. While documentation completeness is intrinsic, the following three definitions of completeness are extrinsic and can only be applied once a research task has been identified.Breadth: a record contains all desired types of data.Some secondary use scenarios require the availability of multiple types of data. EHR-based cohort identification and phenotyping, for example, often utilize some combination of diagnoses, laboratory results, medications, and procedure codes. Quality of care and clinician performance assessment also rely upon the presence of multiple data types within the EHR (the relevant data types vary depending upon clinical area). More broadly, researchers interested in clinical outcomes may require more than one type of data to properly capture the clinical state of patients. In the above cases, therefore, a complete record may be one where a breadth of desired data types is present. It is important to note that the absence of a desired data type in a record does not necessarily indicate a failure in the clinical care process or in the recording process. Rather, it may be that a data type that is desired for research was not relevant from a clinical standpoint, and therefore was not observed.Density: a record contains a specified number or frequency of data points over time.In many secondary use scenarios, EHR data consumers require not only a breadth of data types, but also sufficient numbers and density of data points over time. Some of the phenotyping algorithms developed by the eMERGE Network, for example, rely upon the presence of multiple instances of the same laboratory tests, diagnoses or medications and sometimes specify desired time periods between the recording of these data within the EHR. Clinical trial eligibility criteria, which can be compared to patient records to identify relevant cohorts, also contain complex temporal data specifications, as do EHR data requests submitted by clinical researchers. Breadth and density can be considered complementary, orthogonal dimensions of completeness. A single point of patient data, for example, has breadth and density of one.Predictive: a record contains sufficient information to predict a phenomenon of interest.Our final and most complex definition of EHR data completeness arises when one considers that the overall goal of much research is the ability to predict an outcome. It is possible to train various computational models, some of which being more tolerant of missing data than others, using EHR-derived datasets. Researchers may be interested in predicting, amongst other clinical phenomena, disease status and risk, readmission or mortality. Depending upon the model employed, data needs may be implicit, rather than explicit. The metric for completeness is performance on the task, rather than counts of data points. The data that are required are those that are sufficient to make a prediction.Therefore, it may that two records with different data profiles are both complete according to this definition.
2) ​A health record is a confidential compilation of pertinent facts of an individual's health history, including all past and present medical conditions, illnesses and treatments, with emphasis on the specific events affecting the patient during the current episode of care.Example is EHRs include information like your age, gender,ethnicity, health history, medicines, allergies, immunization status, lab test results, hospital discharge instructions, and billing information.Whereas, A medical chart is a complete record of a patient's key clinical data and medical history, such as demographics, vital signs, diagnoses, medications,treatment plans, progress notes,problems, immunization dates, allergies, radiology images, and laboratory and test results.Example is the data collected includes administrative and demographic information, diagnosis, treatment, prescription drugs, laboratory tests, physiologic monitoring data, hospitalization, patient insurance, etc. Individual organizations such as hospitals or health systems may provide access to internal staff.