Interview Query
~6 hrs
AnalyticsForecasting & Time SeriesBusiness CaseData Viz & InterpretationStatistics

MAPMG Hospital Physician Overload and Discharge Rate Analysis

Investigate whether a high patient census suppresses hospital discharges, using simulated admission and discharge data for 20,000 patients, then estimate the extra patient days to support a staffing cost-benefit analysis.

Mid-atlantic permanente medical groupMid-atlantic permanente medical group
Data AnalystData Scientist
Updated 7 days agoReviewed byIQIQ Team

Overview

The Data Science team at MAPMG is excited to work with innovative students who have strong analytical skills and an interest in learning. As part of the interview process, applicants are asked to work on a realistic problem faced by MAPMG. The data and problem statement presented here are simulations, but they represent the type of problem you may face while working with us.

This problem is not easy. A perfect answer is not expected (nor is there a known perfect answer), but we are interested in seeing how you approach a difficult problem. This problem may require creativity; please don't get discouraged or omit analyses that didn't work: we want to see how you think.

Problem Statement

Two years ago, MAPMG opened a new inpatient wing at the Maryland Hospital for Data Scientists (MHDS). For the first few months, everything was fine: but then, some of the new physicians began reporting that they were having days where they were getting overwhelmed. The hospital is currently staffed with 15 doctors, each of whom is expected to work with about 10 patients per day. Since the average patient census (number of patients in the hospital on a given day) is around 150, some members of the management team are inclined to disregard these complaints.

Physicians and management have asked you to investigate whether there are strong indicators that the physicians are getting overworked, and they've settled on a metric that is important to the bottom line: patient discharges. Some physicians have reported that the discharge process can be very difficult to manage during busy days. This can lead to a spiral where physicians have to triage more and more cases, preventing them from discharging complex patients, but then they have even more patients to deal with the next day.

The lead physician at the hospital would like you to investigate whether the amount of work at the hospital (i.e., the number of patients) affects the discharge rate. If so, management would like to know how large an effect this has, since that will determine how cost-effective it would be to call in additional physicians on busy days.

The physicians have given you the following directives to help guide your analysis:

  • Some patients are relatively simple, and these cases take very little time to discharge; you can expect that patients with only a few days in the hospital won't be affected.
  • The physicians believe that they're really having trouble when they have to deal with 10 or more patients each, and 8 patients each feels like a very manageable number.

Data

You are provided with the file hospital_sim.csv, which contains patient admission and discharge dates for 20,000 patients over a two-year period. Dates are encoded as integers counting the number of days from the hospital's opening.

ColumnDescription
admit_dtAdmission day (integer offset from hospital opening)
disc_dtDischarge day (integer offset from hospital opening)

Unfortunately, the data available does not include information about the patient's health state or the reason for hospitalization.

Tasks

  1. Investigate the relationship between hospital census and discharge rate. Determine whether the number of patients present on a given day affects whether patients are discharged on that day.

  2. Assess the evidence. If you find no evidence of a correlation between busy days and the hospital's discharge rate, demonstrate that your search was exhaustive. If the physicians are correct, provide sufficient evidence to convince skeptical management and physicians.

  3. Estimate the impact. If you find that high census days do suppress discharges, estimate:

    • How many patients are affected by this phenomenon.
    • How many extra days those patients spent in the hospital as a result.

    This estimate is important because management would like to run a cost-benefit analysis for hiring additional physicians on busy days.

Deliverable

Prepare a single PDF containing:

  • An executive summary of your findings.
  • Supporting evidence for your conclusions, including all relevant analyses, charts, and tables.
  • All source code used to perform the analysis, embedded directly in the PDF. The code must be executable and capable of regenerating every result, chart, and table in the document.

Submit exactly one PDF file.

Tools

Please use Python as your primary development and analytical tool.

Download project files