D910 Task 1 Data Analysis and Forecast Example

How Many Nurses for Winter? Two Years of Emergency Department Arrivals Analyzed and Forecast for a Composite Community Hospital's December-to-February Staffing Decision

Student Name

Leavitt School of Health, Western Governors University

D910: Data Analysis for Healthcare Decisions, Task 1

Course Instructor

Month Day, Year

How Many Nurses for Winter? Two Years of Emergency Department Arrivals Analyzed and Forecast for a Composite Community Hospital's December-to-February Staffing Decision

The Decision

The emergency department (ED) at Pine Ridge Community Hospital, a composite 160-bed hospital, schedules nurses six weeks in advance from a staffing grid built on average daily visits. Last winter, the department ran short of nurses on many days, patients waited longer and the number who left before being seen rose. The ED director has asked for an analysis of the past two years of arrival data and a forecast for December through February that she can use to set the winter schedule.

Getting the number right matters for patients as well as budgets. A systematic review of 17 observational studies found that, although the evidence is limited, higher nurse-to-patient ratios in emergency departments were associated with better patient, staff and organizational outcomes, including lower mortality in the one cohort study that measured it and improved waiting times before diagnosis or treatment (Rivet et al., 2026).

Data and Data Quality

The data come from the ED tracking system: one record per registration, with arrival date and time, for October two years ago through last September. Before analysis, each record was checked against three data quality categories: conformance to expected formats, completeness and plausibility (Kahn et al., 2016). Two problems were found. During a registration system outage last March, 212 patients were registered twice, once on paper and once when the system returned; the duplicates were identified by matching name, date of birth and arrival time within 15 minutes, and removed. Another 38 records had no arrival time; they were kept in the monthly counts but excluded from the hour-of-day analysis. No implausible values, such as arrival dates in the future, were found. After cleaning, the data set contains 87,070 visits.

Monthly Pattern and Growth

The table shows monthly arrivals for both years and the change between them.

Two findings stand out. First, volume grew steadily: every month of Year 2 was higher than the same month of Year 1, by between 3.8% and 5.4%, and total visits rose 4.4%. Second, the department has a clear winter peak. In Year 2, December through February averaged 134.4 visits a day, compared with 116.7 a day in June through August, a difference of about 15%. January was the busiest month in both years.

Within the week, Monday is consistently the busiest day, accounting for 16.8% of weekly arrivals, compared with the 14.3% expected if visits were spread evenly across seven days. Within the day, arrivals climb from 8 a.m., peak between 10 a.m. and 2 p.m., and stay high until about 9 p.m. These seasonal and weekly patterns are typical of emergency departments; a study of three hospital EDs confirmed that daily demand is characterized by seasonal and weekly patterns (Jones et al., 2008).

Forecast for December Through February

Because the monthly pattern repeated closely across both years and growth was steady, the forecast uses a seasonal method: each winter month's Year 2 volume, increased by the average annual growth of 4.4%. This method is transparent, easy for managers to check and consistent with research suggesting that forecasts based on calendar patterns are a reasonable approach for daily ED volumes, with more complex models not consistently more accurate (Jones et al., 2008).

The forecast is about 4,270 visits in December, 4,400 in January and 3,970 in February, or about 138 visits a day in December and 142 a day in January and February. Using the lowest and highest growth rates observed in any month, 3.8% and 5.4%, the monthly totals would fall within about 1% of these figures. Day-to-day variation is much larger than that range: last January, daily visits ranged from 118 to 171. For planning, a typical winter Monday should be expected to bring about 167 patients, and one day in ten may exceed 160 even on other weekdays.

Recommendation

The current staffing grid is built on 128 visits a day, close to last year's annual average of about 122 plus a margin. The forecast shows that this grid will be too low on almost every winter day. I recommend three changes. First, raise the base winter grid to 140 visits a day from December 1 through February 28, which under the department's grid adds one registered nurse to each day shift and one to each evening shift. Second, add a further nurse for a twelve-hour midday-to-late shift on Mondays, when volume is highest. Third, place two float nurses on call each weekday, rather than scheduling to the peak every day, because the peak days are predictable in pattern but not in date.

The ED director should compare actual daily arrivals with the forecast each week. If January visits run more than 5% above forecast for two consecutive weeks, the on-call nurses should be moved into the regular schedule.

Limitations

The forecast rests on only two years of data and assumes next winter will follow the same pattern. A severe respiratory virus season could raise volume well above the forecast, and the new urgent care center that opened nearby in August could lower it. The method also does not account for holidays falling on different weekdays from year to year. These limits are why weekly comparison with actual arrivals is part of the recommendation.

Next year's analysis can be stronger. With a third year of data, the department could fit a daily regression model using day of week, month and holiday indicators, the calendar-based approach the research supports, and could add a leading indicator such as the county's weekly count of respiratory illness visits reported by the health department. Tracking forecast error each month would also show whether the method is good enough to keep or should be replaced.

References

Jones, S. S., Thomas, A., Evans, R. S., Welch, S. J., Haug, P. J., & Snow, G. L. (2008). Forecasting daily patient volumes in the emergency department. Academic Emergency Medicine, 15(2), 159-170. https://doi.org/10.1111/j.1553-2712.2007.00032.x

Kahn, M. G., Callahan, T. J., Barnard, J., Bauck, A. E., Brown, J., Davidson, B. N., Estiri, H., Goerg, C., Holve, E., Johnson, S. G., Liaw, S.-T., Hamilton-Lopez, M., Meeker, D., Ong, T. C., Ryan, P., Shang, N., Weiskopf, N. G., Weng, C., Zozus, M. N., & Schilling, L. (2016). A harmonized data quality assessment terminology and framework for the secondary use of electronic health record data. eGEMs, 4(1), Article 18. https://doi.org/10.13063/2327-9214.1244

Rivet, C., Amazouz, H., Gratacap, M., Delannoy, Q., Yordanov, Y., Leblanc, J., & URGE Collaborators. (2026). Association between patient-to-nurse ratios in emergency departments and patient, staff, and organizational outcomes: A systematic review. European Journal of Emergency Medicine, 33(4), 237-247. https://doi.org/10.1097/MEJ.0000000000001350