Retrospective chart reviews.
Data in a retrospective chart review have been collected previously in the process of health care provision (Hess, 2004; Krowchuck, Moore, & Richardson, 1995). In most cases, these data are easily accessible to researchers with approval of an Institutional Review Board (IRB). Human subjects' concerns still exist with this type of research, but the data often can be de-identified and handled by the protections provided by IRBs without obtaining patient consent. In addition, this type of data is less expensive and time-consuming to collect than in many other study designs.
In certain situations, it is difficult or even impossible to collect data in a prospective (as it is happening) manner. Emergencies such as accidents and cardiac arrests cannot be predicted, for example, so retrospective data collection may be the only feasible way a researcher can examine factors of interest. In addition, retrospective studies are important in examining conditions for which a long time exists between exposure to an agent and the disease condition. To study this kind of condition, researchers typically locate patients with and without the condition, and search backwards to see what exposure was experienced by the patient with the condition and by the patient without the condition. This is known as a case-control study (Hess, 2004). Retrospective studies, such as clinical case reports, can be useful for rare occurrences, or when a problem first appears and there is no idea beforehand that a study needs to be conducted.
Retrospective chart review data can serve as useful preliminary findings that assist researchers in planning more complex, prospective studies. The data may suggest important avenues that need further research. For example, Poulin-Tabor and Hyrkas (2008) found no significant difference in length of stay between patients who had an auto-transfusion and patients who did not have an auto-transfusion; however, researchers in another study had found a difference (Munoz et al., 2007). The findings of both studies point to an area of research that needs to be conducted in a prospective manner. Another advantage of this type of study is that it shows researchers how a phenomenon occurred in the normal process of health care without the intervention or process of research (e.g., Hawthorne effect, in which people may change because they are aware of being studied).
Retrospective chart reviews consist first and foremost of data that someone in the past decided to collect regularly for clinical and not research purposes (Krowchuk et al., 1995). Clinicians may not have collected all the data that would be important and necessary to examine the condition adequately under study conditions. In addition, certain key demographic data about the patients may or may not be available in the patient record.
Data from patient health records are known to be incomplete, thus affecting their reliability and validity (Wu & Ashton, 1997). Missing data may require more charts to be reviewed than originally planned in order to have enough records with complete data sets. For example, in the Poulin-Tabor and Hyrkas study, three charts were not used due to missing data. Three out of 151 patient charts is minimal missing data; usually many more records are unusable. Sometimes the data are present but without the detail needed, particularly patient behavioral data (Lyons, Rawai, Yeh, Leon, & Tracy, 2002). In addition, some data may be restricted by the organization.
The process of abstracting data from the record also may produce missing data. Data abstractors may miss data because information has been documented in a different place in some records than in others. In addition, the abstractors may lack clinical expertise in the particular area under study, be tired, or bored with the repetitive nature of the task, and miss data that are present (Aaronson & Burman, 1994).
One of the validity issues with health care record data is the assumption that the data in the health record accurately represent what happens. This will depend on the type of data being collected. For example, quality of care may be difficult to glean from health care records. Researchers have found that retrospective chart data underestimate the quality of care for common outpatient conditions as compared with standardized patient reports (Luck, Peabody, Dresselhaus, Lee, & Glassman, 2000). Records tend to be more accurate with objective services, such as laboratory results and immunization (Eder, Fullerton, Benroth, & Lindsay, 2005), and therefore appropriate for data such as length of stay, weight, drainage, and medications used in the Poulin-Tabor and Hyrkas study.
One of the most important weaknesses is that causation (establishing that one or many factors produced the condition of interest) is difficult to support solely by retrospective studies. Causation may be suggested, but only prospective, randomized studies can support causation fully. Retrospective studies lack the control needed (Hess, 2004). Subjects are not assigned randomly into intervention or control groups. The subjects are in a particular group because of the nature of their condition or the intervention chosen for their care. What this means is that patients may be different at the outset, or were managed differently in other respects (Sheldon, 2001) in ways that were not under the control of the investigators. Results could be due to factors on which data were never collected.
In this type of research, a reader should review how investigators abstracted the data from the chart. A standardized form should be developed based on the variables needed for the study, such as the one mentioned in the Poulin-Tabor and Hyrkas study. Data abstractors should be trained in the use of the form.
As mentioned earlier, a potential exists for variability in the collected data. This variability can occur whether data are collected by one or multiple abstractors (To, Estrabillo, Wang, & Cicutto, 2008). Procedures usually are conducted to ensure intra-rater (same person) and inter-rater (different people) reliability, and should be reported in a research article. For intra-rater reliability, frequently a sample of the charts is re-abstracted by the same person and the results are compared. For interrater reliability, the procedures usually consist of having each abstractor collect data on the same sample of the charts and also compare the data. Statistics such as a kappa statistic may be used to examine agreement. In some studies, investigators even examine the various type of data collected, as some data are abstracted more easily than others. Other procedures are possible, but they should make sense for the particular study.
Despite some of the disadvantages and weaknesses, retrospective chart reviews are useful means to study many types of conditions and can provide useful preliminary findings to support future research. Readers need to be aware of the inherent limitations of any kind of study when reading, and especially when using the findings.
Aaronson, L.S., & Burman, M.E. (1994). Use of health records in research: Reliability and validity issues. Research in Nursing and Health, 17, 67-73.
Eder, C., Fullerton, J., Benroth, R., & Lindsay, S.P. (2005). Pragmatic strategies that enhance the reliability of data abstracted from medical records. Applied Nursing Research, 18, 50-54.
Hess, D.R. (2004). Retrospective studies and chart reviews. Respiratory Care, 49(10), 1171-1174.
Krowchuck, H.V., Moore, M.L., & Richardson, L. (1995). Using health care records as sources of data for research. Journal of Nursing Measurement, 3(1), 3-12.
Luck, J., Peabody, J.W., Dresselhaus, T.R., Lee, M., & Glassman, R (2000). How well does chart abstraction measure quality? A prospective comparison of standardized patients with the medical record. The American Journal of Medicine, 108, 642-649.
Lyons, J.S., Rawai, R, Yeh, I, Leon, S.C., & Tracy, P. (2002). Use of measurement audit in outcomes management. Journal of Behavioral Health Services, 29(1), 75-80.
Munoz, M., Kuhlmorgen, B., Ariza, D., Haro, E., Marroqui, A., & Ramirez, G. (2007). Which patients are more likely to benefit from shed blood salvage after unilateral total knee replacement? An analysis of 581 consecutive procedures. Vox Sanguinis, 92(2), 136-141.
Poulin-Tabor, D., & Hyrkas, K. (2008). An evaluation of postoperative blood salvage and re-transfusion in a total knee arthroplasty patient population: A retrospective study. MEDSURG Nursing, 17(5), 318-322, 330.
Sheldon, T.A., (2001). Biostatistics and study design for evidence-based practice. AACN Clinical Issues: Advanced Practice in Acute and Critical Care 12(4), 546-559.
To, T., Estrabillo, E., Wang, C., & Cicutto, L. (2008). Examining intra-rater and interrater response agreement: A medical chart abstraction study of community-based asthma care program. BMC Medical Research Methodology, 8, 29.
Wu, L., & Ashton, C. M. (1997). Chart review: A need for reappraisal. Evaluation and the Health Professions, 20(2), 146-163.
Lynne M. Connelly, PhD, RN, is an Assistant Professor, University of Kansas, School of Nursing, and Clinical Nurse Researcher, University of Kansas Hospital, Kansas City, KS. She is Research Editor for MEDSURG Nursing.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Research Roundtable|
|Author:||Connelly, Lynne M.|
|Date:||Oct 1, 2008|
|Previous Article:||Evaluation of postoperative blood salvage and re-transfusion in a total knee arthroplasty patient population: a retrospective study.|
|Next Article:||Nurses' stories: discovering essential nursing.|