NSG/577 Week 4: Collecting and Displaying Quality Data, sample paper

Two Points Down Is Not a Trend: Collecting Home Health Hospitalization Data and Displaying It on Run and Control Charts That Separate Signal From Noise

[Student Name]

University of Phoenix

NSG/577: Continuous Quality Monitoring and Outcomes Improvement

Week 4 Assignment

[Instructor Name]

[Date]

The agency, its data and all figures are a composite written for a model paper.

At the last quality meeting, a manager reported with relief that the hospitalization rate had dropped from 19.0% to 17.8%, and another asked what had worked. Nothing had changed in how care was given that quarter. The agency was reading variation that chance alone could explain. This paper sets out how the agency will collect its data and display it so that real change can be distinguished from noise.

Collecting the Data

Hospitalization data come from three sources, reconciled monthly: transfer assessments completed by clinicians when a patient is admitted to a hospital; notifications from the regional health information exchange when a patient registers at a participating hospital; and discharge records. A quality analyst matches the sources for each patient; where they disagree, the analyst checks the visit notes and calls the patient or hospital if needed.

The three process measures are pulled from the visit scheduling system and the electronic record. Each measure has a data dictionary listing the exact fields used, so the calculation can be repeated by anyone.

Checking Data Quality

Before display, three checks are run each month: completeness, the proportion of patients with a start-of-care assessment and a known 60-day status; timeliness, the proportion of transfer assessments completed within two days of the event; and accuracy, a sample of 10 records reviewed by hand against the automated result. Standardized assessment items vary in reliability (O'Connor & Davitt, 2012), so accuracy checks focus on the items used in the risk score.

Why a Control Chart

A bar chart against a target line invites the question of why each month was above or below it, when most months differ for no reason at all. Statistical process control separates two kinds of variation (Benneyan et al., 2003). Common-cause variation is the background scatter that any steady process produces. Special-cause variation is the kind that comes from a real change, good or bad, in how the work is done. A control chart plots the measure over time with a center line at the average and upper and lower control limits, usually three standard deviations from the center.

For proportions such as the hospitalization rate, the appropriate chart is a p-chart. Because the agency admits between 100 and 140 patients a month, the control limits are calculated for each month's sample size; a month with fewer patients has wider limits.

The Agency's Baseline Chart

Using 24 months of data with the Week 2 definition, the average hospitalization rate is 18.1%. For a month with 120 patients, the control limits are approximately 7.5% and 28.7%. All 24 months fall within the limits, and no runs or trends are present. The process is stable, meaning the monthly ups and downs, including the change the manager celebrated, reflect common-cause variation. Improvement will require changing the process, not explaining individual months.

Rules for Detecting Change

The agency will treat any of the following as a signal of special-cause variation: a single point outside the control limits; a run of eight months in a row above, or eight in a row below, the center line; six consecutive points steadily increasing or decreasing; and two of three consecutive points near a control limit on the same side. Benneyan et al. (2003) describe these and similar rules and caution that adding too many rules raises the chance of false alarms. When improvement work begins, a sustained shift below the center line will be the evidence that care has changed.

Monthly and Quarterly Views

Monthly points make the chart responsive, but with about 120 patients each, the limits are wide. The agency will display monthly points for detecting shifts and report quarterly rates, with narrower limits, to the board.

Why Not a Run Chart Alone

A run chart, which plots data over time with a median line, is simpler and was considered. Run charts detect shifts and trends well but do not show how far a single month may stray by chance. Because the board asks about individual quarters, the control limits give a direct answer to whether a given result is unusual. The agency will still use run charts for new measures with fewer than 20 data points, before stable limits can be calculated.

Annotating the Chart

Each time the agency tests a change, the start date will be marked on the chart with a short label. Annotations turn the chart into a record of what was tried and when, so that a later shift can be linked to the change that preceded it. Without them, the agency would be left guessing a year later which of several changes made a difference.

Teaching Staff to Read the Chart

A control chart is unfamiliar to most field staff. Before the first posting, each team will spend 15 minutes of a staff meeting on a single example: the month the manager celebrated, shown inside the limits, with an explanation of why it was not a signal. Short, concrete teaching of this kind is more likely to stick than a written guide, and it prepares staff to recognize a true shift when their own work produces one.

Displaying Process Measures

Process measures are shown on the same page as the outcome, each on its own p-chart. Placing timely initiation, medication reconciliation and front-loaded visits beside hospitalization lets readers see whether process changes precede outcome changes, which Week 5 will test statistically.

Three Audiences, One Data Set

Clinicians see team-level charts for the process measures they control, updated monthly, with a short list of patients whose care missed a measure so they can learn from specific cases. Managers see agency-wide and team-level charts for all measures, with the balancing measures from Week 3 alongside. The board sees quarterly charts of the lead measure and balancing measures, with a one-paragraph interpretation. The same definitions and data feed all three, so the numbers never conflict.

Design Choices

Each chart shows the center line and limits, marks the target as a separate line and labels any signal with a short note of what the team learned. Color is used only to mark signals, not to rate months as good or bad. Axes start at zero for proportions to avoid exaggerating small changes. Donabedian (1988) warned that measures of quality are only as useful as the judgment applied to them, and a display that invites overreaction works against that judgment; the chart's design is meant to slow readers down just enough to ask whether a change is real.

Who Maintains the Charts

The quality analyst updates the charts by the tenth of each month, after the prior month's data have been reconciled. Charts are posted on the shared drive and printed for team meetings, where most field staff will see them. The analyst also keeps a change log recording any correction to past data, so that a revised point is explained rather than silently replaced.

Conclusion

The agency's hospitalization data are collected from three reconciled sources, checked for completeness, timeliness and accuracy and displayed on p-charts that separate common-cause from special-cause variation. The baseline chart shows a stable process averaging 18.1%, so the quarter-to-quarter change that prompted celebration was noise. Clear signal rules, paired process charts and audience-specific views will let the agency recognize real improvement when it comes.

References

Benneyan, J. C., Lloyd, R. C., & Plsek, P. E. (2003). Statistical process control as a tool for research and healthcare improvement. Quality and Safety in Health Care, 12(6), 458-464. https://doi.org/10.1136/qhc.12.6.458

Donabedian, A. (1988). The quality of care: How can it be assessed? JAMA, 260(12), 1743-1748. https://doi.org/10.1001/jama.1988.03410120089033

O'Connor, M., & Davitt, J. K. (2012). The Outcome and Assessment Information Set (OASIS): A review of validity and reliability. Home Health Care Services Quarterly, 31(4), 267-301. https://doi.org/10.1080/01621424.2012.703908