A Consumer Perspective of Health Care and Artificial Intelligence

By Martin Block
Professor Emeritus, Northwestern University

This article examines the consumer response to AI-powered chatbots. We explore this in various settings and then consider their views on privacy. In general, we find that consumer preference for interaction with AI chatbots is lowest in health care compared to other areas such as online shopping. In general, we find that acceptance of AI chatbots is related to the acceptance and usage of other digital services.^[1]

Chatbots

Respondents were asked about their preferred method of communication (a person versus a chatbot) and if they need assistance across five categories. Nearly one-third prefer an AI Chat Bot when engaging in online shopping. The proportion preferring an AI Chat Bot declines for the other categories. Telecommunication and entertainment drops to about 27%, followed by travel, banking, and finance. Healthcare matters dropped to 15% or about half that of online shopping. Note that the focus here on health care is  limited to the consideration of the preferred institutions for non-life-threatening illnesses, such as family doctor’s offices, walk-in clinics, or telemedicine, and the purchase of prescription drugs.

If you need assistance, do you prefer to communicate with?

Categories that involve more personal information tend to have lower AI Chat Bot preferences. Perhaps the lower AI Chat Bot preference for banking relates to the relative complexity involved.

Those that indicate a preference for healthcare AI Chatbots (15.5% of the total) also prefer other forms of contactless relationships with businesses, as shown below. The difference is greatest for contactless checkout and contactless delivery.^[2] 

Which preferred when frequenting businesses?

Telemedicine 

Almost 3 out of 10 (29.5%) respondents have reported using a telemedicine service during the Covid-19 pandemic. Of those that use telemedicine, this proportion rises to 36.5% among those that express a preference for an AI Chat Bot. The consumer understanding of telemedicine varies, with 15.6% reported the last telemedicine event was with their family doctor or Zoom and 26.2% reported they could not remember which website they used.

Respondents were also asked where they prefer to go for non-life-threatening illnesses shows that 10.2% prefer a telemedicine service. This moves to 16.5% among those preferring the AI Chatbot. Overall, the family doctor is preferred.

For non-life-threatening illnesses, what do you prefer to use/go to?

A discriminant analysis shows differences between the five categories (from the above chart) using four predictors age, presence of children in the household, income and being married (Wilks lambda = .857, sig = <.001). The structure matrix is below and shows the dominance of age as a discriminator.

Discriminant Structure Matrix Predicting Preferred Institution

The plot shows that age is the most important discriminator, which is telemedicine preferers are younger than family doctor office preferers. The presence of children shows a slight increase in telemedicine services and a considerable increase in the hospital emergency room. Walk-in clinics, both located in a retailer and not affiliated with a retailer, are somewhat related to telemedicine services.

Respondents were not asked why they prefer telemedicine services but were asked about walk-in clinics. Convenient hours (43.9%) and convenient locations (43.5%) were by far the most important reasons for visiting a walk-in clinic. 

Prescription Drugs

During an average month, the typical household spends $35.73 on prescription drugs. However, only 71.6% of households spend anything, creating a somewhat larger average spend of $49.91 among those who do pay for prescription drugs. Prescription drug costs are commonly subsidized by health insurance plans, leaving the reported cost as out-of-pocket. Costs are highest among those with no health insurance and those through an employer’s plan. This difference may reflect age differences and family size. Those on a government plan report paying the least. The average monthly spend might be interpreted as the amount of healthy activity in the household. Those that prefer the AI ChatBot are more likely to have insurance through an employer or not have health insurance at all. Those with government insurance are considerably less among those favoring AI Chatbots.

Have Health Insurance

The retail format of the prescription drug provider shows drugstores (CVS, Walgreens) as the most preferred provider at 44.2%, discount stores (Walmart, Target) are second at 11.8%, followed by grocery stores (Kroger) at 7.5%. The internet (Amazon) is only 2.9%, but nearly double among those that prefer AI Chatbots at 6.6%. The other category (i.e., the VA) and no preference show relatively elevated levels of average monthly spending. Grocery stores may be slightly misleading since some drugstore brands operate pharmaceutical operations within the store, such as Walgreens in Kroger stores. Drugstores drop considerably among the AI Chatbot preferers to 32.4%, which may indicate a future problem for the retail format if the AI Chatbot preferring group grows in the future. In terms of monthly spending, the internet preferers spend the most by far at $80.79.  Drugstores and grocery stores preferers spend the least.

Retail Format Shop at Most Often Prescription Drugs

The net promoter score (NPS) is the willingness to recommend on a ten-point scale, the net between promoters (the top two choices) and the detractors (bottom six). The membership warehouse (Costco) has the highest score at 40.7, and the internet has the lowest at 10.1. No doubt the membership warehouse is influenced by the price, and the internet may be among the highest users or those with the most health problems. Those with no preference are the least willing to recommend their last prescription drug purchase.

When asked if the cost of health care will affect spending in other areas, almost one-third (31.7%) agreed that it would be likely. There is considerable disparity shown between the AI Chatbot preferers and all others. The AI Chatbot group is nearly twice as likely to see an effect on spending elsewhere (49.2% vs 28.7%).

How likely are your current healthcare costs to affect your spending/saving in other areas?

Privacy

As an indication of privacy concerns, respondents were asked about the use of personal data for advertising.  The largest response is “I don’t like it” at 62.9% overall.  The AI Chatbot preferer, however, is almost four times as likely to say they like it at 39.7%.  The willingness to have the data sold to other advertisers is slightly less favorable.  The overall “I don’t like it” increases to 69.3%.    The AI Chatbot preferer are also somewhat less likely to  say they like it at 34.9%

Do you like it when social media sites take your personal, online, and mobile location data and allow advertisers to use it to send you advertising?

How do you feel about advertisers who buy your personal data to target their advertising to you?

Consumers report taking a variety of steps to protect their online privacy. The steps listed below are not mutually exclusive. The AI Chatbot preferers are slightly less likely to have not taken any steps to protect their privacy, 20.6%, compared to 26.6% of the live person preferers. It is a major concern for both groups with three-fourths of the sample reporting that they have taken steps. The biggest difference between the AI Chatbot preferers is denying permission to track their location followed by posting less on social media.

Steps taken to protect your online privacy

Online privacy is of sufficient concern to generate calls for legislation that prevents the selling of personal and location data. The AI Chatbot preferer is almost twice as likely (18.5% vs 10.7%) to oppose such legislation. However, a clear majority favors such legislation. Online data privacy is a major problem.

Would you like to see legislation enacted that prevents social media sites from selling your personal, online, and mobile location data to advertisers and others?

Summary

We consider consumer use and preferences in health care of chatbots, telemedicine, and online prescription drug purchases and refills. Acceptance of these AI-driven applications is relatively low, in the range of 10% to 15%. The driving forces behind these applications so far are price and convenience. The group favoring telemedicine, for example, are younger which may indicate wider future acceptance as the population ages. Online and data privacy will be an ongoing problem as the majority of consumers favor legislation that restricts sharing or selling of their data. However, those that are users of AI-driven applications see value in the various institutions both having and sharing their data.

[1] The data for this analysis was collected by Prosper Technologies in January 2023 as part of their national monthly Consumer Actions and Intentions Study (n=7,616). The data is collected online, balanced according to age and gender, and includes adults 18 and older.  The margin of error is less than 1.1%.  

[2] We have selected health care chatbots (15.5% indicated by AI Chat Bot) preference as a proxy for artificial intelligence favorability.

Back to AI Edge: Fall 2023