Nudging Teams toward Cost-Effective Health Care: Evidence from a Field Experiment in Surgery

The treatment for both surgeons and nurses will consist of three components. First, at the beginning of each OR shift (morning and afternoon), treated individuals who are scheduled to work the shift will receive a personally customized text message to their mobile phone. The message comprises a general nudge and procedure specific information. The nudge reads “Dear Dr./Nurse X, For the surgeries this morning we remind you to avoid unnecessarily opening disposable equipment”. The procedure-specific information is a list of the planned procedures for the shift and for each procedure a recommended threshold (quantity) of input use for three candidate inputs. The threshold is the median quantity of that input used in the procedure in the 24 months preceding the experiment and approved by the medical professionals on the research team. Second, treated individuals have access to a website that includes a table of thresholds for all candidate inputs. Staff can use this website to check thresholds for inputs that are not included in the message, or when changes in scheduled surgeries occur and no messages are sent. The access to the threshold website is regulated by e-mail and password with access granted only to treated staff. The final component of the treatment is a bi-weekly feedback email, sent to treated surgeons and nurses, containing information on the candidate inputs actually used during the surgeries they conducted over the last 14 days, compared to the recommended amounts.

For details on how we chose candidate inputs, please see the pre-analysis plan.

2022-04-10

2022-10-28

input use (deviations from the recommendation), quality of care

Our main indicator of input use is deviations from the recommendation. Specifically, in addition to computing the median use of each input at the procedure level using baseline data, we also compute the standard deviation. Then, for each operation and input during the experiment, we compute how many standard deviations above (or below) the median input use was.

We use four indicators of quality of care: an indicator for surgery complication, re-surgery in the same department within 7 days, readmission to the same department of the hospital within 7 days, and death within 7 days. The indicator for surgery complication turns on if any of the following seven variables turns on: actual surgery time, transfer to cardiology department or ICU within 24 hours, imaging order after surgery, excess time in recovery ward, blood loss (hemoglobin levels after versus before surgery), kidney failure (creatinine levels after versus before surgery), and surgical site specific infection

See pre-analysis plan

See pre-analysis plan

All nurses and all surgeons in the Operating Rooms department will be invited to participate in the study via individual “introductory meetings” with the Study Coordinators in the month prior to the start of the study. During the meeting, the Study Coordinator will administer a short baseline survey to all participating nurses and doctors, discussed in more detail below. After the baseline surveys are complete, we will conduct the randomization. Half of surgeons and half of nurses will be randomly assigned to the treatment group. This generates four types of teams: those with an untreated surgeon and an untreated nurse, those with just a treated surgeon, those with just a treated nurse, and those where both the surgeon and nurse are treated. The treatment period will last for ~9000 scheduled operations, which we estimate will take 6 months. At the end of the treatment period, we conduct an endline survey with nurses and surgeons.

The randomization will be done in the office by a computer. Details are specified in the pre-analysis plan.

Doctor and nurse level

Yes

110 nurses and 175 surgeons. This may change if some doctors and nurses do not consent to participate, which we will confirm when we complete the baseline survey.

9,000 scheduled operations plus any unscheduled operations that occurred during the time frame. (In case of additional waves of Covid-19, we might need to increase the the volume of surgeries to achieve balance across surgery types.)

We will randomize such that 50% of nurses and 50% of doctors are treated. (This should amount to ~55 treated/untreated nurses, and ~87 treated/untreated surgeons, but the final numbers will depend on how many nurses/surgeons consent in the end.)