C813 Task 1 Health Data Statistics and Ethics Example

Ten Fields From a Discharge Abstract: Classifying Health Data, Choosing the Right Statistics and Using Research Data Ethically

Student Name

Leavitt School of Health, Western Governors University

C813: Healthcare Statistics and Research, Task 1

Course Instructor

Month Day, Year

Ten Fields From a Discharge Abstract: Classifying Health Data, Choosing the Right Statistics and Using Research Data Ethically

Purpose

Before a health information professional calculates anything, they need to know what kind of data they have. The type of data decides which statistics are meaningful, which graphs are honest and which comparisons are misleading. This paper uses ten fields from the discharge abstract of a composite 260-bed hospital to explain discrete and continuous data and the four levels of measurement, matches each level with suitable statistics and displays, and then discusses the ethical obligations that apply when the same data are used in research. The research question behind the example comes from the hospital's quality committee, which wants to study whether patients discharged on weekends are readmitted more often than those discharged on weekdays.

Discrete and Continuous Data

Numerical data are either discrete or continuous. Discrete data take separate, countable values with no values in between, usually whole numbers: the number of prior admissions, the number of medications on a discharge list or the number of days in a stay as the hospital counts them. A patient can have two or three prior admissions but not 2.4. Continuous data can take any value within a range, limited only by the precision of the measuring instrument: body temperature, weight, blood glucose and time from arrival to discharge in minutes. The distinction matters for display. Discrete counts are well shown in bar charts with a bar for each value, while continuous measurements are grouped into intervals and shown in histograms.

Categorical data, such as a patient's discharge disposition, are not numbers at all, even when the hospital stores them as codes. A disposition coded '01' for home and '06' for home health is a label; adding or averaging those codes produces nonsense.

Levels of Measurement Applied

Four levels, or scales, of measurement were described in a classic paper, each allowing more mathematical operations than the one before (Stevens, 1946). Nominal data name categories with no order. Ordinal data have a meaningful order, but the distances between values are not equal or known. Interval data have equal distances between values but no true zero, so differences are meaningful but ratios are not. Ratio data add a true zero to equal spacing, which makes statements like 'twice as long' are meaningful. The table classifies the ten fields.

Matching Statistics and Displays to Data

Nominal data are summarized with counts, percentages and the mode. The share of discharges that occurred on weekends, or the percentage of patients sent home with home health, are the right summaries, displayed as bar charts or, for a small number of categories that add to 100%, a pie chart. The readmission rate for weekend and weekday discharges is a comparison of two percentages, which can be tested for significance with a chi-square test.

Ordinal data are summarized with the median and percentiles, and displayed with bar charts in the order of the categories. Averaging triage levels, for example reporting a mean acuity of 2.7, implies that the distance from level 2 to level 3 equals the distance from level 3 to level 4, which is not true. Pain scores are often averaged in practice, and the hospital should at least report the median and the distribution alongside any mean.

Interval data such as temperature support the mean and standard deviation, and they can be shown in histograms or, over time, in line charts. Ratio data support every statistic, including ratios and coefficients of variation. Length of stay deserves particular care. Its distribution is usually skewed to the right, with most patients staying a few days and a small number staying weeks. In the composite hospital, the mean length of stay is 4.9 days, but the median is 3 days, because a few long stays pull the mean upward. Reporting only the mean would suggest that the typical patient stays nearly five days. The median, together with a box plot or histogram, gives a more honest picture, and the mean can still be reported for financial planning, where total days matter.

Ethics of Using Health Data in Research

The weekend discharge question moves the data from routine reporting into research, and research brings its own obligations. The Belmont Report set out three principles for research with human subjects: respect for persons, which requires informed consent and protection for those with reduced autonomy; beneficence, which requires minimizing harm and maximizing benefit; and justice, which requires that the burdens and benefits of research be fairly distributed (National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research, 1979). Federal regulations built on those principles require that research involving identifiable information be reviewed by an institutional review board (IRB), which decides whether consent is needed or can be waived.

For a retrospective study of discharge abstracts, obtaining consent from thousands of former patients is impractical, and the risk to them is low if the data are protected. The IRB may therefore approve a waiver of consent and, under HIPAA, a waiver of authorization, provided the study could not practicably be done otherwise and privacy safeguards are in place. The alternative is to use data that are no longer identifiable. HIPAA allows de-identification either by removing eighteen specified identifiers, including names, dates more specific than the year and geographic units smaller than a state with limited exceptions, or by an expert's determination that the risk of re-identification is very small (U.S. Department of Health and Human Services, 2012). Because the weekend study needs exact discharge dates, full de-identification would remove the key variable. A limited data set, which may retain dates under a data use agreement, is the better choice.

Ethics also governs how results are reported. The HIM analyst should report every result the study planned, not only the significant ones, describe the limits of the data, and suppress very small counts that could identify individuals. Classifying data correctly is part of this obligation, because a misleading average is a form of inaccurate reporting even when no one intends it.

Conclusion

Correctly classifying health data is the first step in analyzing it. Nominal and ordinal data call for counts, percentages and medians; interval and ratio data allow means and more, with care for skewed measures such as length of stay. When data are used for research, the principles of respect, beneficence and justice, IRB review and HIPAA's rules on authorization and de-identification protect the patients whose records make the research possible. For the HIM professional, accurate statistics and ethical use are two parts of the same responsibility: making sure that what the data are said to show is true.

References

National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research. (1979). The Belmont report: Ethical principles and guidelines for the protection of human subjects of research. U.S. Department of Health, Education, and Welfare. https://www.hhs.gov/ohrp/regulations-and-policy/belmont-report/index.html

Stevens, S. S. (1946). On the theory of scales of measurement. Science, 103(2684), 677-680. https://doi.org/10.1126/science.103.2684.677

U.S. Department of Health and Human Services. (2012). Guidance regarding methods for de-identification of protected health information in accordance with the Health Insurance Portability and Accountability Act (HIPAA) Privacy Rule. https://www.hhs.gov/hipaa/for-professionals/special-topics/de-identification/index.html