HINF 520 Week 1 Data, Information and Knowledge in Health Care Example

Three Answers to One Question: How Many of Our Patients Have Diabetes? Data, Information and Knowledge in a Health System's First Population Report

[Student Name]

University of Phoenix

HINF/520: Data Management and Design in Health Administration

Week 1 Assignment

[Instructor Name]

[Date]

The health system, its counts and its decisions are composites written for a model paper; research findings come from the sources listed.

At a quality committee meeting, a board member of a composite health system with four hospitals and 60 clinics asked what sounded like a simple question: how many of our patients have diabetes, and how many have it poorly controlled? The system's new director of enterprise analytics promised an answer in two weeks. What she learned in those two weeks illustrates the difference between data, information and knowledge, and why that difference matters to administrators.

Data, Information and Knowledge

Data are raw, unorganized facts: a laboratory value of 9.4, a diagnosis code E11.65, a prescription for metformin. Information is data organized and given context so it answers a question: a patient's A1c was 9.4% in March, above the target. Knowledge is understanding built from information and experience: patients in certain clinics with A1c above 9% rarely receive follow-up within three months. A review of the hierarchy that runs from data to information, knowledge and wisdom found that authors define the levels differently but agree that each level adds meaning, context and usefulness to the one below (Rowley, 2007). Wisdom, at the top, is the judgment to act well on knowledge.

Three Sources, Three Counts

The director's team drew data for patients seen in the past two years from three sources. Billing records with a type 2 diabetes code in the E11 category identified 41,200 patients. Active problem lists with a diabetes entry identified 36,900. Laboratory results showing an A1c of 6.5% or higher, or prescriptions for diabetes medications other than metformin alone, identified 44,800. The three lists overlapped but were far from identical.

Why the Counts Differ: Billing Codes

Diagnosis codes are assigned for billing and are shaped by documentation, coding rules and payment. A review of how codes are produced described errors that can enter at each step, from the patient's account and the clinician's documentation to the coder's interpretation and the rules of the classification, and urged users to judge code accuracy for their specific purpose (O'Malley et al., 2005). The team found patients coded with diabetes after a single visit in which diabetes was only suspected, and patients with diabetes whose visits were coded only for other problems.

Why the Counts Differ: Problem Lists

Problem lists depend on clinicians keeping them current. The team found patients whose diabetes was managed by a specialist outside the system and never added to the list, and old entries for gestational diabetes that had never been resolved.

Why the Counts Differ: Laboratory and Medication Data

Laboratory and medication data are more objective, but not perfect. Some patients took metformin for prediabetes or polycystic ovary syndrome, and some had a single high A1c during a hospital stay that later normalized. Patients whose tests were done outside the system were missed entirely.

Judging Data Quality

Weiskopf and Weng (2013) surveyed how researchers had judged the fitness of record data for reuse and grouped the concerns into five dimensions: whether the data are complete, whether they are correct, whether sources agree with one another (concordance), whether values are believable (plausibility) and whether they are current. The diabetes counts failed on several: problem lists were incomplete, some codes were incorrect, the three sources lacked concordance, a few A1c values were implausible, such as 45%, from unit errors, and some diagnoses were out of date. The disagreement among the counts was not a failure of the analysis; it was the analysis, showing where the data could and could not be trusted.

Reconciling the Sources

The team built a combined definition: a patient has diabetes if any two of the three sources agree, or if a single A1c of 6.5% or higher is confirmed by a second result or a diabetes medication. To test it, two nurses reviewed 300 randomly selected charts. The combined definition agreed with chart review in 94% of cases, compared with 86% for billing codes alone.

From Information to Knowledge

With a validated definition, the team counted 43,100 patients with diabetes, of whom 11,600 had a most recent A1c above 9%. Organized by clinic, the information showed that poor control was concentrated in eight clinics serving lower-income neighborhoods, where only 38% of patients with an A1c above 9% had a follow-up test within three months, compared with 61% elsewhere. Discussion with clinic managers added context: those clinics had lost two diabetes educators and had long waits for appointments.

From Knowledge to Decision

The board member's question led to action. The system redirected two care managers and a new diabetes educator to the eight clinics, added a standing order for follow-up A1c testing and began a monthly report of patients above 9% without a recent test.

What Administrators Should Take From This

Three lessons emerged. First, the answer to a question depends on the data used to answer it, so every report should state its definition. Second, data quality must be assessed, not assumed, and the dimensions offer a checklist. Third, knowledge requires context from people who understand the work; the numbers alone did not explain why eight clinics lagged.

Data From Outside the System

One gap affected every count: care received elsewhere. About 18% of the system's patients also saw clinicians in other organizations, and their outside A1c results reached the record only when a patient brought a paper copy or a clinician requested it. The regional health information exchange could supply many of those results, but the system received them as documents rather than structured values that analytics could use. Until outside data arrive as structured, coded results, any count of controlled diabetes will understate what patients and their other clinicians have achieved and overstate the gaps.

Who Owns the Definition

The exercise raised a governance question: who decides what counts as diabetes for the system's reports? The quality department, the population health team and the finance office had each been using a different definition for years, producing reports that disagreed in front of executives. The director proposed that the clinical quality committee approve one definition, stored in the data warehouse and used by all departments, with a version number and date so that changes are visible. That step, more than any software, prevents the next board meeting from receiving three answers again.

The Role of Information Systems

The exercise also showed why the health system needs better data management: a shared definition stored in one place, data quality checks that run automatically and a way to bring in outside laboratory results. Those needs will shape the diabetes data project described in the next papers.

Limits

The combined definition still misses patients whose care happens entirely outside the system and may include a few with resolved diabetes after weight loss surgery. The team documented these limits in the report so users understand them.

Conclusion

A board member's simple question produced three different answers, because data from billing, problem lists and laboratories each captured diabetes differently. Applying the hierarchy from data to knowledge, judging quality on five dimensions and validating a combined definition turned conflicting data into reliable information and then into knowledge that changed care in eight clinics.

References

O'Malley, K. J., Cook, K. F., Price, M. D., Wildes, K. R., Hurdle, J. F., & Ashton, C. M. (2005). Measuring diagnoses: ICD code accuracy. Health Services Research, 40(5 Pt 2), 1620-1639. https://doi.org/10.1111/j.1475-6773.2005.00444.x

Rowley, J. (2007). The wisdom hierarchy: Representations of the DIKW hierarchy. Journal of Information Science, 33(2), 163-180. https://doi.org/10.1177/0165551506070706

Weiskopf, N. G., & Weng, C. (2013). Methods and dimensions of electronic health record data quality assessment: Enabling reuse for clinical research. Journal of the American Medical Informatics Association, 20(1), 144-151. https://doi.org/10.1136/amiajnl-2011-000681