Risk and Personalized Support for Patients and Families: A Field Experiment in a Major Chinese Hospital

2023-10-11

2024-10-11

Risky decisions of the family measured by the non-compliance degree against physicians.

Both health outcomes of the pregnancy women and the newborn, and a series of answers in the follow-up surveys.

We collaborate with a leading hospital in China to conduct a field experiment, involving high-risk pregnancy women participating in the Multi-discipline Treatment (MDT) and their companions (if any). Using a between-subject design, we vary the content and recipient of information at the family level, including pregnant women and their companions, across treatments before MDT. In the control group, both subjects in the family receive basic introduction about the hospital and MDT. In the treatment groups, pregnant women receive additional disease information related to them in Disease group or both disease information and emotional support in Emotion group. In Family group, pregnant women receive the same information as in Emotion group, while companions also receive additional disease and emotional information. To ensure exposure to the information, we send extra information based on assigned treatments every two weeks, for a total of four times after MDT. Moreover, we use a baseline survey to measure the degree of risk aversion, ambiguity aversion, rationality score of both subjects in the family, as well as the coordination level between them before the information intervention. Following each information distribution after MDT, we attach a survey to elicit subjects’ risk perception, emotion, and their coordination level to examine the potential mechanisms of our intervention.

1. Participant
We focus on high-risk pregnancy women enrolled in MDT, suffering from various pregnancy comorbidities such as diabetes and Systemic Lupus Erythematosus (SLE). By participating in MDT, they can receive professional and comprehensive suggestions from physicians across different departments. Due to the significance of MDT, they are always accompanied by their relatives, primarily their husbands. Thus, it provides us a unique opportunity to treat the pregnant women and their companions simultaneously. Hereafter, we refer to the family as subjects including the pregnant women and, at most, one of their accompanying companions in the MDT. If a woman has more than one companion, we include them as our subjects following this order: husband, mother, and other relatives. If a woman has no companions, the family only consists of one subject.

2. Treatment
Our between-subjects experiment consists of one control group and three treatment groups (Disease, Emotion and Family). The randomization unit is the family.
1) In the control group, both subjects in the family receive basic information before MDT, including the introduction about the hospital, obstetrical department, MDT procedure and physicians in the MDT. After MDT, only pregnant women receive general pregnancy precautions every two weeks, including aspects such as sports, diet, and medication.
2) In the Disease group, in addition to the information in the control group, pregnant women also receive personalized disease information about their conditions before MDT, along with disease-related precautions instead of general ones in the control group after MDT. Their companions receive the same information as in the control group.
3) In the Emotion group, in addition to the information in the Disease group, pregnant women also receive general emotional support before MDT, and personalized emotional support according to their pregnant weeks, ages and risk preferences after MDT. All emotional support messages are generated by Large Language Model (LLM). Their companions receive the same information as in the control group.
4) In the Family group, pregnant women receive the same information as in the Emotion group. However, their companions receive personalized disease information related to pregnant women, and tips to make them provide emotional support to pregnant women both before and after MDT.

3. Baseline survey.
We distribute the baseline survey for both the pregnant women and their companions before the information intervention on the day of MDT. We measure risk preference using certainty equivalence, ambiguity preference using Ellsberg’s two-color urns, rationality score using standard budgetary task, and coordination level between them using adapted coordination game.

4. Follow-up survey.
In the biweekly follow-up surveys, we aim to examine the mechanisms of our intervention. Therefore, we elicit subjects’ risk perceptions, emotions, and coordination level between the couple.

5. Hypothesis
We test the following hypotheses through our field experiment:
Hypothesis 1: Families are less likely to make risky decisions in all treatments (Disease, Emotion and Family), compared to the control group.
Hypothesis 2: Families are less likely to make risky decisions in the Emotion group, where pregnant women receive additional emotional support, compared to the Disease group.
Hypothesis 3: Families are less likely to make risky decisions in the Family groups, where companions receive additional disease information and emotional support tips, compared to the Emotion group.

Additionally, we examine the following hypotheses using our baseline survey:
Hypothesis 4: The risk aversion of both subjects in the family is negatively associated with the likelihood of making risky decisions.
Hypothesis 5: The ambiguity aversion of both subjects in the family is negatively associated with the likelihood of making risky decisions.
Hypothesis 6: The rationality score of both subjects in the family is negatively associated with the likelihood of making risky decisions.
Hypothesis 7: As the coordination level decreases, the associations between risk aversion, ambiguity aversion, and rationality score of the subject with low bargaining power in the family and the likelihood of making risky decisions decreases.

Randomization is conduct by a computer. For each week, we randomly assign the new subjects into different groups to minimize the difference of balanced variables across groups.

Randomization unit is the family, including the high-risk pregnant woman and, at most, one of her companions (husband, mother or other relatives).

No

Around 320 families.

The unit of analysis is a family. Thus the planned number of observations is 320.

We will have four groups (control, Diease, Emotion and Family), and each group contains 80 families.