Surveying Healthcare Workers in Nursing Roles

The first stage of the experiment estimates how nursing staff tenure and experience may affect the quality of healthcare provider care and patient outcomes. The study will elicit beliefs from real healthcare workers about the importance of on-the-job experience for delivering high-quality care. Participants are randomly assigned with equal probability into two conditions regarding the experience level of a hypothetical nurse. We measure how beliefs about relative quality of care received by patients change when we exogenously vary the amount of experience a hypothetical nurse has.

The second stage estimates beliefs about how healthcare provider financial conditions affect worker departures and reservation wages. Participants are randomized into five conditions about the financial health of a hypothetical facility. We measure how required wages and willingness to work at a hypothetical facility change when we exogenously vary information about the firm's financial health.

2022-11-04

2023-02-01

In the first stage of the experiment, the primary outcomes are the time a hypothetical nurse takes to complete a task and the likelihood a hypothetical nurse makes a mistake or a patient has a bad outcome. We truncate both of these values such that answers below 1 are replaced with the value 1 instead. This corresponds to 1 minute to complete a task or a 1% chance of a mistake or bad outcome.
In the second stage of the experiment, the primary outcomes are the likelihood the participant would search for another job if working at a hypothetical facility, the reservation wage the participant would require to stay at that facility, and the participant's beliefs about the fraction of workers at the hypothetical facility that would voluntarily leave.

Our secondary outcomes are answers to direct questions about how a facility's financial health affects workers' willingness to work at that facility and questions about how staff tenure impacts quality of care.

Participants are randomly assigned with equal probability into two conditions regarding the experience level of a hypothetical nurse. We measure how beliefs about relative quality of care received by that nurse's patients change when we exogenously vary the amount of experience the hypothetical nurse has.

Next, participants are randomly assigned into five conditions describing the financial health of a hypothetical facility. We measure how each participant's willingness to work at the facility and reservation wage required to work at that facility change with the financial health of the facility.

We conduct our experiment using a Qualtrics study. Qualtrics features a computerized randomization method, which we use to randomize participants into the treatment conditions described above.

We randomize at the participant level.

No

We exclude individuals that fail attention checks. Additionally, we exclude individuals that have never worked in the healthcare industry at a skilled nursing facility as an RN, LPN, or CNA. We also exclude participants who do any of the following: (i) report a probability above 100, (ii) report a time to complete a task above 4 hours, (iii) report a historical wage greater than $100 per hour, or (iv) give answers that are self contradictory as explained in our analysis plan. We will continue to run the experiment until either (i) we have at least 240 completed surveys that meet these criteria or (ii) we have continuously run our survey through February 1, 2023 without reaching 240 complete surveys that meet these criteria. Because of our filters, and because of the nature of our online-survey vendor, we cannot precisely control the number of participants or end date of our experiment.

We exclude individuals that fail attention checks. Additionally, we exclude individuals that have never worked in the healthcare industry at a skilled nursing facility as an RN, LPN, or CNA. We also exclude participants who do any of the following: (i) report a probability above 100, (ii) report a time to complete a task above 4 hours, (iii) report a historical wage greater than $100 per hour, or (iv) give answers that are self contradictory as explained in our analysis plan. We will continue to run the experiment until either (i) we have at least 240 completed surveys that meet these criteria or (ii) we have continuously run our survey through February 1, 2023 without reaching 240 complete surveys that meet these criteria. Because of our filters, and because of the nature of our online-survey vendor, we cannot precisely control the number of participants or end date of our experiment.

We exclude individuals that fail attention checks. Additionally, we exclude individuals that have never worked in the healthcare industry at a skilled nursing facility as an RN, LPN, or CNA. We also exclude participants who do any of the following: (i) report a probability above 100, (ii) report a time to complete a task above 4 hours, (iii) report a historical wage greater than $100 per hour, or (iv) give answers that are self contradictory as explained in our analysis plan. We will continue to run the experiment until either (i) we have at least 240 completed surveys that meet these criteria or (ii) we have continuously run our survey through February 1, 2023 without reaching 240 complete surveys that meet these criteria. Because of our filters, and because of the nature of our online-survey vendor, we cannot precisely control the number of participants or end date of our experiment.

In the first stage, participants are randomized, with equal probability, into two groups regarding the experience level of a hypothetical nurse. If we have exactly 240 participants, this implies ~120 participants per group.

In the second stage, participants are randomized into a control group or one of four treatment groups. Each group sees different information about a hypothetical facility. We randomly assign one-third of participants to Control. We randomly assign the remaining participants with equal probability to one of the four treatment groups. We thus allocate participants as follows: Control, one-third (~80 participants); Treatment 1, one-sixth (~40 participants), Treatment 2, one-sixth (~40 participants), Treatment 3 , one-sixth (~40 participants), Treatment 4, one-sixth (~40 participants). Most of our analysis combines treatments 1 and 2 together into one condition and treatments 3 and 4 together in one condition. See our analysis plan for details. Note that because of our filters, and because of the nature of our online-survey vendor, we cannot precisely control the number of participants or end date of our experiment. In particular, it is very likely we will not end up with exactly 240 participants so these distributions are approximate.

NurseSurvey_AnalysisPlan_Nov2.pdf